Olivier's adventures in Wonderland

pyexiv2 0.3.2 released

Olivier Tilloy — Mon, 24 Oct 2011 19:39:00 +0200

I am pleased to announce that pyexiv2 0.3.2 was released today.

This is a brown-paper-bag release that fixes bug #880659 (Regression: pyexiv2 0.3.1 doesn’t work with Python 2.6).

The source tarball is available on Launchpad, as well as a win32 installer and packages for Ubuntu will soon be available in pyexiv2 developers’ PPA.

As usual, your feedback is much appreciated.

pyexiv2 0.3.1 released

Olivier Tilloy — Sun, 23 Oct 2011 20:37:00 +0200

I am pleased to announce that pyexiv2 0.3.1 was released today.

This is a maintenance release.

The highlights of this release are:

Compiles and tested against the latest libexiv2 (0.22)
Updated windows dependencies (iconv 1.14, libexiv2 0.22, python 2.7.2, boost 1.47.0)

The source tarball is available on Launchpad, as well as a win32 installer and packages for Ubuntu will soon be available in pyexiv2 developers' PPA.

Many thanks to everyone who contributed with bug reports, patches, testing and suggestions. As usual, your feedback is much appreciated.

This bug needs fixing…

Olivier Tilloy — Thu, 28 Jul 2011 14:44:00 +0200

Confort digital

Olivier Tilloy — Sun, 20 Feb 2011 19:10:00 +0100

Savez-vous pourquoi les rangées de touches sur nos claviers d’ordinateurs sont décalées les unes par rapport aux autres? À l’origine des claviers on trouve les machines à écrire mécaniques : les lettres gravées sur des blocs de métal étaient fixées au bout d’une tige rigide, et venaient imprimer le caractère correspondant sur la feuille de papier lorsqu’elles étaient activées. C’est donc une contrainte mécanique qui a dicté l’agencement des rangées de touches sur les premiers claviers. Si les rangées avaient été alignées les unes par rapport aux autres, comment toutes les tiges métalliques auraient-elles cohabité sans se gêner? Avec l’électronique et les claviers tels que nous les connaissons aujourd’hui, cette contrainte a entièrement disparu, mais les rangées de touches décalées sont restées, probablement à l’origine pour ne pas contrarier les habitudes des dactylographes et autres virtuoses de la machine à écrire.

Saviez-vous que la disposition des lettres et autres caractères sur nos claviers (le célèbre AZERTY, équivalent national de l’américain QWERTY), n’obéit à aucune logique scientifique ou linguistique, mais bien mécanique, encore une fois à cause des machines à écrire? À l’origine, il est probable que les tous premiers claviers aient utilisé l’ordre alphabétique pour la disposition des lettres. Mais avec la popularisation des machines à écrire, les dactylographes professionnel(le)s tapant de plus en plus vite, un problème est apparu : les tiges métalliques se coinçaient les unes avec les autres si elles étaient activées de manière successive trop rapidement, ce qui devait arriver souvent pour quelqu’un capable de taper "à l’aveugle", sans regarder le clavier. L’idée brillante des fabricants de machines à écrire de l’époque a été de réorganiser les lettres sur le clavier afin que les séquences de lettres les plus fréquentes dans la langue anglaise ne soient pas consécutives sur le clavier (ce qui réduisait les symptômes du problème sans pour autant le résoudre…). De même que pour les rangées décalées, avec l’électronique et les claviers tels que nous les connaissons aujourd’hui, le problème original a complètement disparu : vous pouvez taper aussi vite que vous le souhaitez sur un clavier, pas de risque d’enrayer la mécanique! Et pourtant, le QWERTY et l’AZERTY sont restés. Dur de changer nos habitudes, surtout quand ça implique remettre en question notre manière d’interagir avec les ordinateurs, omniprésents dans notre société.

Pourtant, des gens ont fait ce travail de remise en question depuis longtemps, et nous proposent des solutions pour améliorer notre confort de frappe. J’ai découvert le bépo et le typematrix il y a un peu plus de quatre ans, mais ça n’est que l’an dernier que j’ai décidé de franchir le pas. Je suis informaticien, et en cette qualité, je passe une grande partie de la journée devant mon écran, les doigts sur le clavier. Je suis donc particulièrement exposé aux risques de troubles musculosquelettiques qui affectent les utilisateurs de claviers, comme le syndrôme du canal carpien et autres tendinites. J’ai donc décidé de remettre en question mes outils de travail afin d’améliorer mon confort au quotidien et de donner une chance à mes doigts.

TypeMatrix™ est un fabricant américain qui produit des claviers ergonomiques. Le concept de base est simple : les rangées de touches décalées n’ont plus aucune raison d’être depuis plusieurs décennies, elles sont donc alignées, et les touches des différentes rangées forment ainsi des colonnes parfaitement verticales (d’où l’idée de "matrice"). D’autre part, certaines touches d’utilisation très fréquente (entrée et retour arrière), qu’on trouve habituellement à l’extrême droite du clavier, contraignant l’auriculaire à d’incessantes contorsions, sont situées sur une colonne au centre qui divise le clavier en deux, accessibles par les deux mains, et naturellement à portée de l’index, qui est un doigt fort.

Le bépo quant à lui est une disposition alternative des caractères sur le clavier. Contrairement au QWERTY et à ses dérivés, sa composition obéit à un travail rigoureux de placement des lettres en fonction de plusieurs critères, dont les principaux sont :

fréquence d’apparition dans la langue (français en l’occurence)
fréquence de groupes de lettres consécutifs dans la langue (digrammes et trigrammes)
équilibre de la charge de travail pour chaque main

Ainsi, l’objectif de la disposition n’est plus de pallier aux insuffisances mécaniques d’une machine vieille de 200 ans, mais bien d’optimiser le confort de l’utilisateur. Et ça se ressent au quotidien. Contrairement à une idée reçue assez populaire parmi les informaticiens, il ne s’agit pas d’atteindre des vitesses de frappe vertigineuses (bien qu’une vitesse de frappe accrue soit une possible conséquence d’un plus grand confort). Pour ma part je ne tape pas (ou guère) plus vite qu’avant, en revanche je fais beaucoup moins de fautes de frappe, je tape maintenant vraiment "à l’aveugle" (mon clavier est d’ailleurs vierge), en utilisant mes dix doigts, et je ne souffre plus de tensions dans les doigts et les poignets à la fin d’une longue journée de travail.

Apprendre à utiliser un nouveau clavier et une nouvelle disposition n’est pas chose aisée, ça nécessite de la volonté. C’est un investissement qui peut sembler un peu fou au départ. Au delà du prix du clavier, c’est surtout du temps, de l’assiduité et de la patience qui sont nécessaires pour ré-apprendre à taper de zéro. En quelque sorte, c’est comme apprendre à parler une langue étrangère ou à jouer d’un instrument de musique. Les maladresses linguistiques et les fausses notes sont une étape obligée qu’on laisse bien vite derrière soi. Et avec la maîtrise vient le plaisir.

C’est aussi un investissement sur le long terme. Dans quel état seront mes doigts dans 50 ans? Et ceux de mes collègues qui se battent contre leur clavier tous les jours? Et puis il faut bien l’avouer, l’anti-conformiste qui sommeille en moi retire une grande satisfaction de cette évolution.

Si vous avez lu jusque-là et que la disposition bépo a éveillé votre curiosité, je vous recommande vivement d’aller faire un tour sur le wiki du projet. Il est extrêmement bien documenté et animé par une communauté active et très accueillante.

FOSDEM 2011

Olivier Tilloy — Wed, 09 Feb 2011 20:00:00 +0100

Last week-end I attended the eleventh edition of FOSDEM in Brussels. Like the previous editions, it was simply amazing. Great minds, anonymous hackers and free software enthusiasts, in total more than 5000 visitors over the course of two very intense days! Volunteers who saw to the organisation did an impressive job at running the whole event smoothly, thank you very much guys!

Unfortunately two days is a very short time frame for so many talks, discussions and Belgian beers, and this year again the whole thing was over before I had time to realize. And the lack of sleep didn’t help either…

The event traditionally kicked off with the infamous Friday Beer Event. Picture hundreds of hackers gathered in a bar in the centre of Brussels, cheap and tasty Belgian beer… That’s right, the logical consequence is a bad hangover on Saturday morning. Luckily Ugo had a load of painkillers that did wonders to my head and saved my day. As last year, the bar was really packed and I didn’t move around much, still I managed to bump into Matthias Saou, a former colleague at Fluendo, and I also met a bunch of Igalians, some of whom I had met last year in the same place.

Saturday

This year I had decided to travel very light, and I intentionally didn’t carry any laptop or netbook, the idea being to really make the most of the talks. A decent phone able to connect to the network for schedule updates and meeting workmates would have helped, though… So I went around carrying a notepad and a pen, taking detailed notes of the presentations I had carefully selected in a small printed version of the schedule. And I must say that overall I’m not disapointed by the quality of the presentations I attended.

Eben Moglen’s opening keynote was inspiring. He talked about the power of ordinary people to change things, and how freedom may look like something messy, destructive in the short term, but is salvaging in the long term. He went on to highlight the enormous political importance of social networks, and how highly dangerous to use the current solutions are, because they are too centralized, and because business and profit will always prevail over technology and freedom. And he said we need to fix this, by creating federated social networks. We need to operate under the assumption that the network is untrusted and untrustworthy. He finally announced the creation of the Freedom Box Foundation.

Chris Lattner then presented the LLVM project. Interesting to know that there are people out there who are working hard to improve our most fundamental tools, compilers and the rest of the toolchain.

After a quick lunch break, I went to the Mozilla devroom to see Tristan Nitot give a quick speech about the foundation in 2011 (or was it 20011 as he said?). The new marketing motto of Mozilla for Firefox seems to be: «We report to no-one but you».

I then moved to the nearby cross-distro devroom to listen to Thomas Weber, a debian developer, explain how to be a good downstream. By means of a list of common-sense advice, he advocated for two-way pro-active communication between packagers and upstream developers. Communication and empathy are, he said, the keys to a successful relationship, and will avoid most of the issues that usually arise. In an ideal world, upstream developers shouldn’t have to deal with distribution-specific stuff (it’s great if they are willing to do so though), and packagers need to understand that upstreams may have different deadlines and schedules than that of their favourite distro.

The next talk in the same room was given by Jared Smith, current Fedora Project Leader, and was entitled «Swimming upstream, or how distributions help open source communities». He developed on the river analogy (upstream, downstream, salmons and other apparently unrelated topics), to make a point on how distributions actually create communities (who said Ubuntu?), how discussion, as long as it’s respectful and civilized, is a healthy thing (who said Code of Conduct?), and to ask what artificial barriers we are putting in people’s way, which reminded me very much of the ongoing efforts of Canonical’s community team to make sure the processes in place help people rather than get in their way. He concluded on why we are doing all this: at the end of the day, we all live downstream.

I then hurried back to the main lecture hall for Lennart Poettering’s introduction to systemd, a drop-in replacement for sysvinit. I had already read his introductory paper, so I knew what it was about, and the prospect of starting all processes in a completely parallel manner is really exciting, but what proved really interesting was the update on the status of the project, and the questions that were asked at the end of the talk. Having worked recently on optimizing the boot sequence of a netbook for a project within Canonical OEM services, I can’t wait to see the benefits of such an init system in action.

I then went to a presentation of Firefox 4 for developers by Tristan Nitot. I’m not much of a web developer myself, but Firefox’s latest performance improvements, together with full support for HTML5 and CSS3, promises an exciting future, and the demos he showcased were truly impressive. Too bad the presentation suffered from obvious lack of rehearsing.

The last discussion I attended on Saturday was about XMPP and federated social networking. After a presentation of buddycloud, a rather technical discussion ensued about implementation details. It seems several groups of people are currently experimenting with various approaches, and they hope to converge towards a common standard at some point. Good to see a real example of constructive discussion and community efforts.

And with that Saturday at FOSDEM was over, which I took advantage of to take a quick nap before going for dinner with my friend and workmate Ugo.

Sunday

Sunday started early for me with a talk in the accessibility devroom entitled «How does a blind person see a computer with free solutions», by visually-impaired Jean-Philippe Mengual. I am particularly interested in this topic in the context of Unity in Ubuntu, its currently poor support of a11y, and the total lack thereof in Unity 2d at the moment. The GNOME desktop and Ubuntu are often praised by enthusiasts as being quite accessible for people with physical or cognitive impairments, and there is indeed a collection of tools dedicated to making the desktop usable for them, but in practice eSpeak, the default text-to-speech synthesizer, was deemed not usable by the presenter. Ouch. Good commercial, proprietary solutions exist, but they are very expensive (> 2000€ for IBM’s text-to-speech tool). The good news is that projects aiming at producing better free tools exist, that there is a whole team in Debian dedicated to a11y, that people developing GNOME 3 are working hard to enhance a11y, and that every single of those efforts is a step forward towards enabling the impaired to fully get rid of proprietary solutions. The presenter, a mere user himself, urged developers to have a11y in mind when writing software, and gave a series of common-sense advice to achieve that.

I then went to the embedded devroom to see Loïc Minier talk about Linaro. That was a very instructive presentation. Until then, to me Linaro was a big word which I knew was related to Linux on ARM, and that my employer is investing heavily in, but little more really. Loïc delivered a high-level overview of the goals of the project, its actors and its current status, and then dived in the details of the current and future tasks of various of the teams. With the increasing number of projects for ARM platforms we are seeing at Canonical’s OEM Services, this effort makes a lot of sense and I’m glad I’m now slightly less ignorant on the topic.

I stayed in the same room for a talk entitled «Qt tales from the embedded trenches». The abstract was promising, and it could have been interesting, especially since I’m currently up to the neck in Qt and QML with Unity 2d, but the speaker was trying too hard to be fun, and showing too much boring code on screen. Don’t show me code, give me food for thought!

I left this last one early, which gave me the opportunity to meet Bertrand Lorentz, of Banshee fame, who I had known last year at UDS in Brussels. It seems exciting stuff is planned for future versions of Banshee, and that includes DVD support and better video classification (unfortunately but quite predictably no contribution from Fluendo in sight).

I then headed towards the main lecture hall again for a practical go programming crash course. Andrew Gerrand, who works at Google in Sydney, took the audience through the exercise of writing a URL shortener in Go, with the goal of demonstrating some of the neat features of the language, such as defers, interface types and built-in concurrency with goroutines and channels. Definitely interesting, although maybe not of immediate use to me, but I’ll keep an eye on it.

I met with my colleague Vincenzo Di Somma for lunch, which gave us the occasion of exchanging about our respective ongoing projects.

I then headed towards the main lecture hall again for Michael Meeks’ «The Wonderful World of LibreOffice, or how we will fix your office suite». He is certainly a great speaker, as I already had the opportunity to witness in the past. He probably digressed too long on how evil contributor agreements are, especially since I had already read a lot of this on his blog. Not that I entirely disagree with him, but at some point I even wondered whether he was going to get back to the main topic, LibreOffice. He eventually did, and the future looks bright for our favourite office suite: lots of new contributions from the community, new features, but also less sexy yet essential tasks, like tedious housekeeping to clean up the code base from its cruft, including thousands of lines of comments in German! In his words, «we wanna be a cool, vendor-neutral, diverse, linux-like community».

Back to the cross-distro devroom for a talk on automated testing (in this case of OpenSUSE). Scripted tests are run daily on new images in virtual machines, and the success of the tests are determined by taking screenshots of regions of the screen, and comparing with known-good expected results. Additionally, the system produces a fast-forward video (one frame captured every half-second) which allows an operator to later review what happened in case of failure. This allows to spot regressions early in a fully automated way, although at the moment it still involves a lot of initial work to set it up. The solution seems to be generic enough to allow testing other OSes: the presenter mentioned Debian, Fedora, OpenBSD, and even Windows! This is work in progress and could use a lot of improvements, but the concept is interesting. Results are presented in a web UI. I wonder how well this complements/competes with mago and checkbox on Ubuntu.

I stayed in the same room for the following talk, entitled «One source to rule all binaries». Sascha Peilicke presented the OpenSUSE build service that allows to build packages and custom distributions out of spec files. He demonstrated how from one seed the system is able to build packages for a large number of distributions, including Ubuntu, Fedora, Mandriva and Debian. Definitely cool and impressive, yet subject to lots of potential issues inherent to differences between distributions. Different package naming across distros is one example, custom patches another obvious one. Probably a very good tool for upstream developers who want to experiment with packaging their software for cheap.

And with that it was already time to catch a bus to the train station and then a train to the airport. I missed the closing keynote («How kernel development goes wrong and why you should be a part of it anyway» by Jonathan Corbet), but luckily it was recorded, which will allow me to watch it later.

Bye-bye Brussels, see you next year!

Going to FOSDEM

Olivier Tilloy — Sat, 22 Jan 2011 12:57:00 +0100

It’s that time of the year again… February is around the corner, and with it the FOSDEM event.

I’ll be in Brussels from the 4th to the 6th of February. Arrival planned Friday afternoon, in time for the Friday Beer Event, of course.

This year I’ll try to be more selective on the talks I attend, and make the most of my time there by meeting interesting people and engaging in hallway discussions.

Looking forward to it!

pyexiv2 0.3.0 released

Olivier Tilloy — Fri, 31 Dec 2010 18:10:00 +0100

I am pleased to announce that pyexiv2 0.3.0 was released today.

This series remains fully backward compatible with its predecessor, the 0.2 series, which should ease the transition away from the antiquated 0.1 series.

The highlights of this release are:

Compiles and tested (on linux and windows) against libexiv2 0.19, 0.20, 0.21
ImageMetadata implements the collections.MutableMapping interface
Consistent API across all types of tags to access the value(s)
Read/write access to the EXIF thumbnail
Decode and encode EXIF comments according to the specified charset
API to (un)register custom XMP namespaces
API to get, set and delete the (optional) IPTC charset
Added pickling support to tags
Use fractions.Fraction when available in the standard library (Python ≥ 2.6)

The source tarball is available on Launchpad, as well as a win32 installer and packages for Ubuntu (Lucid, Maverick, Natty) in the pyexiv2-developers' PPA.

Many thanks to everyone who contributed with bug reports, patches, testing and suggestions. As usual, your feedback is much appreciated.

And a happy new year!

Where is my bike?

Olivier Tilloy — Wed, 11 Aug 2010 19:58:00 +0200

Almost two years ago I wrote a little Python script that retrieves availability information for the bicing network in Barcelona. My grand plans to use that script to optimize my morning routine didn’t quite see the light, but the script itself seems to have proved useful to others.

Earlier this year, Eskerda contacted me to let me know that he was interested in writing an application for android to assist bicing users in cycling the city efficiently. As bicing.cat’s servers are dead slow, hitting them directly from each running instance of the application was not an option, so he came up with the idea of a static resource hosted on Google App Engine and updated regularly, using my script.

The result is OpenBicing, a cool application for android. I don’t own an android phone myself, so I couldn’t test it, but it looks really fancy, I particularly like the radius mode. Bonus point, it’s Free Software (Apache license).

And as if it wasn’t good enough, Eskerda set to work to support more bike sharing networks with the same model. OpenSevici and OpenVélib are two standalone applications that respectively cover the networks of Sevilla and Paris.

In an effort to gather them all in a single application in an extensible way, CityBikes was born. Android users and cyclers, go check it out, and don’t forget to send your feedback to the author!

Note to self: consider buying a new phone…

Hello Canonical

Olivier Tilloy — Mon, 05 Jul 2010 23:09:00 +0200

Today is my first day at Canonical, and I am thrilled to be joining a company that for the past 6 years has led the way and contributed so much to spreading Free Software and delivering Ubuntu to the masses.

I discovered Ubuntu back in 2005, when I first installed the Breezy Badger on my laptop, and I haven't looked back since then! Today I am given the tremendous opportunity to contribute first-hand to this effort within the OEM Services group, and I am really looking forward to working with all the awesome people I had the chance to meet in Brussels at last UDS, and of course all those I haven't met yet!

pyexiv2 0.2.2 released

Olivier Tilloy — Thu, 27 May 2010 12:29:00 +0200

I'm happy to announce that pyexiv2 0.2.2, codename "Holiday", was released today.

This is a maintenance release that fixes two memory leaks, optimizes the use of the underlying libexiv2 (expect performance improvements), improves the API documentation, restores access to the image comments (was a regression from the 0.1 series) and adds an optional parameter to preserve timestamps when writing metadata.

The source tarball and a Windows installer (compiled and tested against Python 2.6.5) are already available, and packages for Ubuntu 10.04 will follow shortly in pyexiv2 developers' PPA.

As usual, feedback, suggestions and bug reports are welcome!