Axel Hecht Mozilla in Your Language

June 6, 2007

Not just incubating

Filed under: L10n,Mozilla — Axel Hecht @ 3:46 am

I have written about the incubator before, it’s a tool that should help localizations newly in CVS to work towards release. It’s based on buildbot, and I whacked it pretty hard recently.

The result can be seen on the l10n server again. What’s new?

  • Incubator is now L10n Trunk, I’m running my tools on all localizations working on the trunk now. That’s both localizations moving towards fx3 and new ones.
  • It’s using the python implementation of CompareLocales. It does check more, and it’s output could potentially suck less. Red again means items missing, orange is obsolete items, green is good.
  • I’m exposing stats on localizations. You’ll find a 90% translated in the output box of the comparison step. More below.
  • There’s one column for trunk, and one for incubators. You can tell which build is yours by looking at the description of the comparison step.
  • Technically, it doesn’t sink the l10n server anymore on incubator check-ins.
  • Only incubator builds do langpacks right now, and only if the comparison ain’t red. There’s a download link labeled [D] on the upload step that takes you there.

So what’s the stats? First of all, I know that “translated” isn’t the right term, but it’s as wrong as anything else, so wth. What I’m doing is, I’m counting all entries in DTDs and properties files which key does not match [kK]ey. Then I’m comparing those values to the en-US values and then do a 100*changed/(changed + unchanged). There are still a bunch of values that don’t need to be changed to be localized, so I’m not worried that we’re not at 100% anywhere. I do think that the differences between the high 70s and the 90 are real, though. I’m confident that removing the keys made those stats comparable mostly, as it did fix Japanese.

Why the redo? As indicated above, the first implementation didn’t scale, so I needed a bit more buildbot-ism. I didn’t feel good about adding new languages to the incubator just for load reasons. That’s solved. The amount of languages leading to numbers of columns turned out to be bad, too.

What the redo? I’ll make follow-up post on that, this one is long enough.

June 5, 2007

2.0.0.4 languages

Filed under: L10n,Mozilla — Axel Hecht @ 7:50 am

There are a few changes in our language support that came with Firefox 2.0.0.4. We have officially released Afrikaans and Belarusian. Congratulations to both teams. If you know someone that should or could use those versions and doesn’t, please point them at the download page, as those links aren’t that easy to find otherwise.

We would like to see as many testers as possible for our beta releases, too. You can find download links to Georgian, Kurdish, and Romanian on the release notes. Those builds are only tested by the localization teams so far and may still contain stop-ship bugs. We need your help to find out.

Feedback on all of these is welcome on bugzilla, or on their respective feedback forums if you’re more familiar with those.

June 4, 2007

LinuxTag 2007 wrap-up

Filed under: Mozilla — Axel Hecht @ 9:32 am

LinuxTag 2007 was interesting, in fact, more interesting than I anticipated. I cherry-picked a few talks from the program on Friday and Saturday.

Crowding effects: How money influences open source projects and its contributors, by Matthias Stürmer. I already blogged about this one a bit. The slides are up by now. Matthias is organizing the OpenExpo in Zürich in September, and we’re invited. I’m waiting for an invitation on English, still.

Best Practices for Open Source Projects, by Ross Turk. Didn’t see all of his presentation, but I did come in time for a picture and to catch that he’s anticipating to see sf.net cooperate with external projects to make their tooling better. They apparently want to improve their site to get beyond a download redirector, too.

Das Phänomen OpenOffice.org – alles Individualisten oder Freaks? by Jacqueline Rahemipour. Not sure if the target audience for that talk was there. Most notably, she explicitly turned down attempts to target increasing female contributions to OpenOffice.org, despite females being a minority there, as elsewhere in the open source community. She wouldn’t see any climate or somesuch making it hard for women to contribute. These notes translated from German spoken word into written English, so take it with a grain of salt. But the core message was “Women in Open Source special project? No thanks.”

The Best of Both Worlds – Open Source und Marketing by Valerie Hoh. The talk was aimed at development leads that haven’t looked into marketing yet, and wouldn’t just because of the name. She introduced a two definitions of marketing, one of which was along the lines of “plan an exchange of goods or services for the benefits of both parties”, which led her to point out that software needs to work for the user, not the developer, and that the developer gets more users and more cookies in return for more users. She had room for one positive example, don’t hurt the web. That’s her wallpaper now :-). I was able to detail on more of our marketing approaches in the following discussion. Later that day, I pointed here at foxkeh.com. As painful as that is to german eyes, it works for Japanese.

I had to report back some dog food bugs on the native OSX OpenOffice.org build I got on Friday on Saturday morning, so I didn’t really get a whole lot of Nicolas Vandenberghe’s Semantic Web, Wiki and Mashup: how they can all work together. He was presenting ITerating.com, a wiki-based software guide. I wasn’t intending to go there for all the RDFness in it, but for some ideas on how to merge wiki editing and more or less structured information. Sadly, I didn’t get a whole lot more than “we’re using RDF on a laid back path, without all the fancyness on top of triples”. They do reification, though, if I got that right. But no OWL or SPARQL or somesuch. They did get asked to open the source of their app, and they’re going to spend some thinking on that.

Grüße aus der Gruft: JavaScript in Zeiten des Web 2.0 by Christian Wenz. This talk was an interesting mix of anecdotal stories, history of fun, grief and fun of using js on the web without blaming Brendan half as much as he does himself at dinner tables. Christian promised to give his left hand for Firebug and was pleased to hear that we’re investigating including it in fx3. I didn’t poke him for a comment on js1.5+ -> 2, as he was focussing on the web.

GNU/Linux stack Localization to right-to- left languages: Persian Language Experience. Mohammad Khansari and Narges Zali’s talk was cancelled, sadly.

Create cross- platform web applications with XUL by Carola Kummert and Arne Blankerts. You probably know Arne as TheSeer from #developers, I had no clue he was in Germany. Hamburg, to be precise. I met Bernd Mielke before that talk, and we both scared the holy cow out of Arne. The first time that folks in the audience raised their hands on the “Are there any core Mozilla hackers in the room?” question. Harharhar, pirate day. The talk was mostly focused on more abstract pros and cons like XBL, better layout, templates for huge datasets. They also differentiated remote XUL vs installed extensions or xulrunner applications. No technicalities on the xml markup level. They mentioned RDF a few more times than it seems to be fashionable these days. The good news is, they’re working on a book. The bad news is, they asked me to proof-read it. They even had some XUL dark matter than I didn’t run across yet, took a screenshot of that.

Besides the actual talks, I did a bunch of networking. I hang around at the OpenOffice.org booth for a bit again, and dug into a few issues in Firefox. I did receive a bunch of compliments to forward to the Thunderbird team for tb2, too. Not to mention, I did wear my tb shirt on Friday, go Scott! The reactions I received when talking to people throughout on Mozilla were positive. I received some questions on structure and cash, in particular as I commented about our giving program on Friday morning, bascially the first thing I did after buying a ticket. I even answered questions about debian and iceweasel to the full satisfaction of those that asked, though I didn’t run into any debian maintainers, AFAICT. The idea of Firefox as a brand wasn’t alienating people at all. When it comes down to community, product, and marketing, I haven’t received any doubts or disagreement, it was a comfortable atmosphere of “Mozilla knows where they’re going and they’re good”.

Carlo approached me on psyced, which I admittedly didn’t get. twentyafterfour is working on a Firefox implementation. I invited to them to do more blogging about their stuff and get it syndicated on planet, and maybe prepare a brownbag to present at the office. It may be a protocol to distribute data real time to multiple recipients, something that artists at myspace.com would be interested in, or taking chat a step further, and both. Sounds confused? I am. I blame myself for not finding the step between network protocols and who cares about yet another multi-protocol chat. Carlo was sure it’s not a dupe of AllPeers nor really p2p, but that those guys should talk.

In general, I can’t really judge on the claims that LinuxTag would be the Europe’s Biggest Trade Fair for Free Software, Linux and Open Source, I would estimate the crowd at FOSDEM to outnumber the crowd on LinuxTag, both this year and last. Even if I consider that Wednesday and Thursday are rumored to have been better and if I multiply the crowd at LinuxTag by 4 (FOSDEM crowd stays for both days, mostly, LinuxTag is likely more one day per visitor), FOSDEM feels bigger. It does feel more technical, too.

Oh, did I mention pictures? Here, of course.

June 1, 2007

Crowding out at LinuxTag

Filed under: Mozilla — Axel Hecht @ 3:37 am

I listened to a talk by Matthias Stürmer on crowding out effects in communities by monetary incentives. Gonna be interesting to have Seth and Matthias look at the Mozilla Community giving program. The good news is, I feel that we’re on the right track here from what I heard so far.

May 31, 2007

LinuxTag 2007, Berlin

Filed under: Mozilla — Axel Hecht @ 5:42 am

I’ll be spending some time at LinuxTag 2007, here in Berlin, Friday and Saturday. There seem to be a few interesting talks on the program, including some XUL dark matter.

I’ll be hitting the exhibition floor, too.

May 14, 2007

Re: platform

Filed under: Mozilla — Axel Hecht @ 5:25 am

Firstly, architectures with one platform give you grass on rails, not rubies. Opting for still-imagery here, take a look at the Amerika Bahnhof. I pondered taking half a day off and take some pictures from Berlin’s Lehrter Bahnhof (or, Hauptbahnhof). Anybody who’s been there knows that it’s a maze of platforms, aiming at different needs, operated by different parties (there are at least 3 different operators for local public transportation in Berlin). That doesn’t count all those independent parties running shops and restaurants.

Mozilla is a train station, and it comes with a plethora of platforms. Platforms as in horizontal stuff to take off from. NSPR, NSS, Gecko (undefined in all levels of –with[-out]-technology), libxul, xulrunner, toolkit applications. Those are platforms for programming. I’d call Firefox rather a platform for web experience, not strictly one for programming, I see non-programming values in Firefox that form a distinct kind of platform, like privacy, security, good defaults, user choice. Then there’s spreadfirefox, which is a platform to launch marketing projects. Not everything that is essential to a train station ecosystem is a platform, btw, and there are plenty of those in Mozilla, too.

Despite the fact that it took me me only a few minutes to write down some punch words for some of our platforms, Mozilla doesn’t do a good job about defining which stuff we consider to be a platform or not. That doesn’t mean that platforms need to be done, or that MoCo has to commit to get them done. They may actually be as much of a Baustelle as anything. Yet, defining them and giving them names makes them easier to find, and to fix, even. I consider the non-programming platforms vital for the train station ecosystem, too.

May 9, 2007

9:30

Filed under: Mozilla — Axel Hecht @ 2:22 am

Firefox Flicks works, listen to the cover version of Wheee in wet feets in plone at minute 9:30.

May 8, 2007

Calling Toronto

Filed under: Mozilla — Axel Hecht @ 1:32 am

Hrm. How many folks do we have in the Toronto office exactly? Just wondering.

April 30, 2007

The incubator

Filed under: L10n,Mozilla — Axel Hecht @ 3:21 pm

When working on new localizations of Firefox and Thunderbird, it’s pretty tricky to get all the tooling right and to distribute your work for testing. In addition, there’s no real build environment that tells me how far particular localizations are. That changed, enter the incubator. Incubator runs in buildbot for the most part, and calls into the Mozilla build system for the real work. I don’t repack builds, as these are really for early testing, I do generate language packs against the last stable release, though. I’m looking forward to the app-agnostic branching for 2.0.0.4, here, too, then the language packs should have a real chance to work against Thunderbird, too. For now, check out the language packs. I’m repackaging the l10n trunk against en-US branch, so we can get lively engineering work on the side of the localizers against a stable and popular target. Thus incubator.

The whole thing runs on the l10n server, that is, master, slave and download location. Thanks to Reed Loden for helping me with the set up there.

This is still work in progress, I hope to see our new localizations pound that thing with check-ins to get some testing and experience.

What works:

  • Bonsai poller. I whacked that poor creature into oblivion, I need it to support multiple subsets of repositories on different bonsai servers and branches.
  • CVS checkout of that same data.
  • Did I say same data? Yep, I actually have one data structure to create both the change source and the source build steps

What works not:

  • Any kind of load balancing or fail over does not work. I do want to stick to the plan to only have the source once per slave, but the concept of “get the source on all my slaves, and then distribute the repacks on them as they come” doesn’t seem to exist in buildbot. Partially, you could get there with scheduler.Dependent, but I guess that would stall the complete build if a slave dies.
  • Adding a download link to the waterfall display.
  • Fast update, only check-out what bonsai poller found.
  • Dependency checking in repackaging. In the real world, one should repackage all locales only for changes to en-US, and for changes in l10n only repackage affected localizations. This is somewhat mid-hanging fruit, I guess I could add a .fileIsImportant check to scheduler.Dependent.

As for the Waterfall display, I’m less sure than before that there is a nice way to display all of our locales. Not sure how bad that is, there seem to be ways to select subsets of the waterfall, for example just the Albanian browser repack.

If you happen to be one of the localizers on there, give bonsai 10-20 minutes to pick up your changes. If you’d like to be on there, and you’re not, poke me.

April 6, 2007

side by side

Filed under: Mozilla — Axel Hecht @ 2:05 am

I have two morning routines, going through my newsgroups, and through my list of comics. Today, side by side, the Mozilla 2 discussion on .platform and today’s strip on sinfest.

« Newer PostsOlder Posts »

Powered by WordPress