• The People: All the awesome folks were there, had great discussions and learned a bunch of new things.
• The Event: Pretty good, presentations mostly technical, BoFs I attended were more or less fruitful, visionary talks did have plans with them, some flames.
• The Partaes: Hotel open terraces rock. Boat triped rocked (thanks Collabora!). Lots of fun! (except the one with cheap beer..)
• The City: Fantastic. Istanbul and the Bosporus are beutiful, pretty European feeling, its people were very friendly, weather was fine.

I took a bunch of photos from the event and the City — check them out at dimitris.glezos.com/photos/events/guadec08/.

I gave a short talk on Transifex. Two points I wanted to stress were the following. One, translations are important for any project that would like to increase its userbase with little sweat (you code for English and let translators help you reach out to a bigger audience). You create for the 5% of the Earth’s population, plus the other 95%. And two, that GNOME’s concerns for the upcoming move to distribute VCSs and translations could become less of a problem once we deploy Transifex in GNOME’s infrastructure.

From the discussions I had, it seems that a lot of people are anxious in setting Transifex somewhere centrally for any open source project, hosted anywhere, to benefit from. There are some maintainers who stick around svn.gnome.org, just for the sake of translations. That’s something that shouldn’t happen — developers shouldn’t be forced by translations to use inferior tools that affect their productivity negatively. Hackers tend to get frustrated quite easily when they don’t have the tools they need. :-/

It also happens that companies like Red Hat, Novell and Nokia, among others, which contribute extensively in open source, spend a lot of money in transferring translations around their downstream community, their translation contractors and upstream. I’ve actually heard the phrase “shitloads of money” at GUADEC. Which translates in many cases in either development time (since contractors often don’t know how to talk with upstream) or just in no sharing of the translations with upstream at all.

Lately I’ve more or less made it my personal goal is to tackle these challenges. Translations should be easy. Dead-simple easy. For both developers (documenters, webmasters) as well as translators.

Can we make it a click away?

Στιγμιότυπα από το GNOME 2.20

<self-talking> Άραγε πόσο δύσκολο θα είναι να προστεθεί υποστήριξη στο meld για να ανοίγει diff αρχεία; Θα βοηθούσε πολύ αν μπορούσαμε να δούμε γραφικά τις αλλαγές που κάνει ο κάθε συντελεστής σε ένα αρχείο… </self-talking>

Photos

Some random pics I shot during my stay. Next GUADEC at Istanbul, I’ll probably save more free time, the weather will most likely be better, and the city will definitely have more interesting stuff to put in the frame.

Our hotel, Etap	J5 explaining XO
Doc Searls keynote

Other related posts:

• The GUADEC diaries — chapter 1
• The GUADEC diaries — chapter 2

• Had a meeting with Danilo Šegan (Данило Шеган) of GNOME L10n fame and Carlos Perelló Marín of Ubuntu, Launchpad/Rosetta fame, and discussed handling of upstream translations. It’s a pretty important issue which everyone would like to see tackled, but there are many technical details that need addressing, like for example, downstream patches that alter the module’s translatable strings (the need only downstream translation, requires coding), security (sigh), and coordination with/review from upstream translation teams (if they exist).

  Danilo and I talked also about translate.fedoraproject.org, damned lies and Transifex. Both of us would like to keep the codebases close to each other, but without breaking or enforcing anything. This is basically my responsibility, but transifex adds a bunch of stuff on DL’s object model, and requires the data to be updateable (ie. XML files are problematic), so we should solve these in DL too. A port to TurboGears will make the project more maintainable.

• Talked a bit with Colin Walters and Bryan Clark about Mugshot’s i18n. There are a bunch of different technologies in the project, so it’s not just a matter of a patch submitted. The Python web pages might be the easier of the three targets (other two are the server and the client).

• Had a cool discussion with a Fedora OpenOffice.org hacker/maintainer about the way OOo handles locales and languages. There are many requests to have OOo override the session’s locale, which imply to have OOo behave differently than all the other desktop apps. This will only increase the confusion and if the problem is solved, it will only be so for OOo. Also, OOo spellchecker ideas about fine-grained language detection are again something that ideally would be solved desktop-wide and not just in OOo. Having the “keyboard layout” define what the language of the text written is, again isn’t a one-to-one relationship; this could work, but only if it was just a suggestion to a guessing engine, and not a true binding between the two. Doing something that makes sense is important.

Update, 2 hours later:

• After my talk, David Glassey from Debian approached me for discussing whether it’s possible to use Transifex in Debian. The answer is definitely! — the whole idea behind transifex is to give the ability to any translator community to contribute directly to a remote VCS, thus bridging the gap between the translation commuinities. The more downstream communities get connected, the better for everyone: more language coverage, higher quality.

  Of course, each bridging attempt brings the possibility of conflicts. People have different opinions on how a phrase should be presented in a different language, and so this is an issue that might come up at some point. Let’s hope a discussion will always bring an answer, make everyone happy, and make the new gate work constructively. :-)

  Talking about gates, Transifex could also be used by an individual translator on his workstation, as a gate to all the projects he is submtting translations to. One can set it up with his SSH keys, get presented with all the projects he has access to, and use it to commit contributions instead of the command-line VCS tools. For those who want to quickly do something without any strange VCS voodoo, they can simply grab the PO file from the statistics page and commit it through transifex in two clicks.

Other related posts:

• The GUADEC diaries — chapter 1
• The GUADEC diaries — chapter 3

The place is packed with all sort of hackers, but mostly GNOME ones (obviously). Wherever you turn your face, you see a familiar head from planet.gnome.org and even some from planet.fedoraproject.org. I attended (all) three talks yesterday, and all of them were interesting.

After the last lecture was finished, we went with some friends for a nice dinner at a french restaurant and returned to the hotel. Entering the lobby one word came to my mind: Pandemonium. The lobby was full of people talking to each other, or hacking on their laptops. Apparently the hotel’s wifi wasn’t prepared for this usage and most people received a dreadful “connection slots are full”. And it’s still the “warmup” period of the conference. Talk about people loving what they do.

Had an hour-long interview with a researcher studying innovation in open source communities. Some questions were interesting, or at least the thinking to give an answer to them was. Of course, the classic Q came up “why do people contribute to open source”, which always makes people think hard, and I guess I gave the classic answer “for a variety of reasons, it depends on the person, but I guess it is a little bit of everything.”

Got the chance to talk with some Ubuntu maintainers about translations in Rosetta not being pushed upstream. A more in-depth discussion should take place with the Rosetta maintainers, who happens to also be here, so more news about this later on.

Other related posts:

• The GUADEC diaries — chapter 2
• The GUADEC diaries — chapter 3