Have you ever been a part of one of these releases?  Here’s a bit of how it works…

On September 10, Axel made the first call for participation in localizing FF 3.1 Beta 1.  Time went by as our localizers prepared their localizations, dealing with new complexities like migrating all their prior efforts to the new code repository, translating newly added strings, and learning how to use Hg.  We answered questions along the way and helped when prompted.

On September 29, Axel made another announcement about Firefox 3.1, this time calling for our teams to “opt-in” to the official release process for the first beta.  We were down to a few days left and people needed to start sending us their work.

By looking at the L10n dashboard (click on “Fx31x” in the Tree section, and the “35″ in the Result section), we were able to see the locales that were closest to finishing up, but hadn’t quite gotten into the process.  Stas and I started emailing teams directly who were close.  We compiled the teams who had unbroken, green localized builds of Firefox, but who had yet to opt-in to the release process.  Then, we gathered all those who had broken, red builds because they had missing strings.  I emailed each team with updates and asked them to opt-in if possible.

At this point, we started to see the locales roll in.  One can’t help but feel a bit of pride and a little thrill as our localizers stand up to the challenge.  The l10n-drivers sit back and watch.

Then, my mind starts to drift to a galaxy far, far away….

I’m always reminded of this scene in Star Wars. (30 seconds in)

In the clip, all wings report in…red 10, 7, 3, 6, 9, 2, 11, and 5…one-by-one with some excitement and pride.  As we approach code freeze, localizers will chime in much the same fashion.

• “fr locale is opting in, patchset a7c003554f46″
• “please add sk too, changeset f3fb0018a3b”
• “hu locale is opting in, patchset 279d4db75211″
• “es-AR opting-in, changeset e35e14d53ef0″
• “ru locale is opting in, patchset a86d6547da37″

and so on…

And, just like the Jedi master Obie-wan, Axel will drop in a whisper with sage advice along the way…

(best read with Obie-wan whisper)

“…that build has missing strings…”

“…if you get the changeset into my queue, you’re fine, but I wouldn’t fall over if I missed out on this…”

“…you can land, and we’ll tag and release with the given changesets….”

When code finally froze (like a tauntaun on the planet Hoth), we had 36 locales* participating!  I believe that’s the most we’ve had in the first beta of a new version of the Firefox browser.

Congratulations to the localizers!  It’s truly an accomplishment that shows the commitment and dedication of our community.  I’m proud to be a part of it.

* see the comment field…

Mostly, it means we have allocated more resources to our l10n build infrastructure with less custom setup to maintain.  We have several identical machines working together to repack l10n builds, while in the past we repacked them one build at a time in one giant pack…see where we are going here? 

Ultimately, this will allow us to better service localizers with build resources.  We can do things like “one last repack” for those localizers who might have that need.  In the past, we lined up all builds in a specific order and started the process as one big batch.  If your build was in there, well, you’d have to wait until next time if something happened.  Now, we have a pool of identical machines working in unison to service the l10n-build queue, grabbing the next build to be done as fast as the machines can.  And, everything will happen much, much faster (perhaps 7x), since we have multiple machines working for us.

For the time being, the “new builds” are in the usual nightly directory here:

ftp://ftp.mozilla.org/pub/firefox/nightly/latest-mozilla1.9.0-l10n/

If you want to view “the old way”, this URL takes you to the builds that are generated the traditional way:

ftp://ftp.mozilla.org/pub/firefox/nightly/old-l10n/

If you see any problems with the new nightlies, before you file a bug, can you check to see if the same problem occurs with the nightlies that are created in the traditional way?  Functionally, these should be identical, but it will be helpful if we can see if the problem is in the new system or in both.

This post from Armen explains more. Please read.

John O’Duinn will post later today or tomorrow with more details.  We are going to run these two systems in parallel for one week.  If there are no complaints, we will shut down the traditional system and only use the new system.  Please send along any questions if you have them.

This quarter, I think we’ve started to make some progress on those questions.  Here’s what I have noticed so far:

L10n dashboard

I won’t spend too much time here because I’ve blogged about it before, but this tool will only become better and better as time goes on.  It is the one tool I look at consistently to remind me what localizations are translating Firefox, who might need to be contacted, and how we are doing as a community upcoming releases.  If you haven’t checked it out yet, please do so.

Weekly Meetings

One thing we added recently is a weekly, Tuesday bug triage call.  In those meetings, Axel uses the l10n-dashboard and Bugzilla to review which localizations are awaiting responses from us or might need a “poke”.  He then gives good recommendations to each of us for next steps.  This is when Axel is playing quarterback for our team and passing out assignments for each member.  Perhaps some of you have noticed that we are trying to proactively get people to start localizing Firefox 3.1.  We’ve been pinging people in bugmail to see if they are ready to move their translations from cvs into Hg.  The whole team meets by phone every Tuesday at 10 AM Mountain View time to discuss who we should ping and which member on the team needs to do what.

Tricks of the trade

I blogged last week about changes to the l10n team.  Here are some additions to our process from the new team members.

Stas has created a Firefox 3 l10n release trackers bug and then shows the dependency tree inside bugzilla to help us track each language we are trying to get into upcoming releases.  It’s a simple change allows us to not let anything slip through the cracks.  With Axel’s l10n dashboard and little things like this blocker bug, we’ll continually track who in our community might need some help or hear from us in bug- or email.

In addition, Gandalf is working closely with me to reach out to some of the newer localization teams in the queue.  We try to gather the status of these teams and see where they might need help in navigating through the early stages of our process.  I often send people to these pages that dicsuss things necessary for starting a new localization.  (Those links discuss Mercurial and starting a new localizaiton).  One specific example on how we have helped, just today, Gandalf and Axel prepared a zip file that presented all the string changes from Firefox 1 to Firefox 3.1 for a localizaiton team from Africa who had stopped its work right around the launch of Firefox 1.  That team hopes to translate and get into the 3.1 release cycle.

Responding to requests

With these tools and team members, I think we are doing a pretty good job at responding to requests.  We’ve scheduled IRC meetings with a lot of teams, including all of the teams that are launching in Firefox 3.0.2.  I’ve also considered scheduling a monthly call for all localizers to dial into.  It would be a time for locales to ask questions and for the team at Mozilla to give a semi-regular in-person update.  This is just an idea, but might be worth a try.  Thoughts?

Adding languages

We added 8 new languages with the upcoming release of Firefox 3.0.2.  By the release of Firefox 3.0.3, we hope to add 5 more.  That’s 13 more languages in the span of two minor updates.  The challenge going forward may be keeping the pipeline filled with languages and making sure we reach out to the locales that have been working away for a long time.

I am up for more suggestions from the community on how we can improve.  Please just comment or email me.

Simon Paquet (IRC nick sipaq) will be leading up nearly all communication about localization efforts for Thunderbird.  Technological questions will also be fielded by Simon and his team, but Axel and the l10n-drivers will remain a very strong support network for any questions that need help.  Nothing changes in how localizers should expect to communicate though…please point questions to the dev-l10n mailing list.

Zbigniew Braniecki (IRC nick gandalf) has joined the l10n-drivers team and will be helping with several things, including developing Silme; supporting localizers by providing code reviews, creating patches, and checking in code; building community with tools like the Community Pack; and being another outlet to ping with all questions.  Gandalf has a long history as a localizer and understands deeply the Mozilla process.  He should be a great compliment to Axel’s hard work and knowledge of just about everything Mozilla l10n.

Staś Małolepszy (IRC nick stas) also joined the l10n-drivers team.  He will be taking over all web services and productization elements of the localization process for Firefox.  Additionally, he will be supporting localizers with code review, creating patches, and checking in code.  He’s also running Mozilla Community Survey program, so we hope to use his talents there to continue to learn more and more about our community.  Staś will also support Pascal Chevrel and our community with the localization of web parts that are included in Firefox.  Staś had his beginnings as a localizer and deeply understands the Mozilla process.  He’ll help with our efforts to make everything run as best and as transparent as possible.

Adrian Kalla has joined Mozilla for a few months as a localization intern hacking on Silme with Gandalf and Axel.  Like Gandalf and Staś, Adrian is a localizer who has a great grasp of our localization process.  We’re excited to have him around for the next few months.

Mic Berman (IRC nick mic) will be moving on from Mozilla.  We are so thankful for all her contributions in helping to guide fully localized releases for Firefox over the past couple years.   Please do send along any thanks and praises to Mic if you’d like.  We are thankful for her service to the Mozilla community.  Thanks, Mic!

As always, please ping us with questions.  Firefox 3.1 Beta is approaching quickly and we are eager to help all localizers with questions.  We want to have a fully localized beta, so if you are a localizer, please let us know what you need to make that happen. 

Gerv and I have both blogged a bit about how Mozilla tries to measure the impact that its community has by localizing Firefox for Web users across the globe.   Specifically, we look at both the number of localizations and the percent of the world’s Internet population covered by Firefox with our translations.  We don’t really measure by country boundaries.

But, to see how we rated against this Chrome metric, Pascal ran an analysis to see how many countries a user could find Firefox.  He started by looking at the total number of countries listed by the U.N. Then, he looked at the Wikipedia entry for each country to find the official language spoken.  By that count, Firefox is in 161 countries.

If we were to add that to our list of metrics, for the upcoming release of Firefox 3.0.2, we can provide the following about Firefox:

• Firefox is present in 161 countries
• Firefox is localized in 57 languages officially shipped across three platforms
• Firefox has 93.1%* of the world’s Internet population covered

I guess my first question that comes up when I think about coverage in terms of countries:  what happens to languages that do not map to geopolitical boundaries?  Guys like Erdal from our Kurdish localization probably have some thoughts on that.  What about India?  Do the people in Andhra Pradesh think that India can be represented as a country if Telugu is not a localized option for their browser?

Admittedly, I think our metrics are a better measurement (though slightly incomplete…see below) of the impact of a localization effort.  It would be nice to open this conversation to the guys at IE, Safari, Chrome, Opera, Flock, etc.  We have a lot of open discussions about security, performance, and standards.  How are these other browsers looking at localization metrics?   Did the Chrome team use the same methodology we did above to get the number of countries?  What are the other metrics that help illustrate impact?  Perhaps it’s wishful thinking, but I’d like to learn more from the other localization teams.

*  I’ve mentioned before how we get this percentage and that our model is a bit incomplete, but here is how we think about our metrics:

Our goal is to count the number of people in the world speaking different languages who have access to the Internet, irrespective of country boundary, and then see where Firefox localizations provide coverage.  Our premise is that some countries (like the U.S.) have people whose native language may not be the official language of that country.  Therefore, we’ve set up a matrix where across the top we list most languages spoken in the world.  We have a column that lists all countries.  Then, we do as much research as possible to find out how many people in each country speak another language “at home”.  The assumption is that people who speak a different language at home would also like to browse the Web in those languages.  In the case of the U.S., one might expect to see its population speaking English, Spanish, Arabic, Chinese, and so on.  The biggest challenge is getting accurate data of who speaks what languages in each country.

Pike resolved bug 450655 which updates shipped-locales for the new beta versions that will be going out in Firefox 3.0.2.  We will be adding the following (I’ve linked to Wikipedia descriptions of each language if you are unfamiliar):

• Bengali (bn-IN)
• Hindi (hi-IN)
• Kannada (kn)
• Marathi (mr)
• Telugu (te)
• Galician (gl)
• Icelandic (is)
• Thai (th)

I visited Wikipedia page to find the low-end number of people in the world who speak these languages.  By adding these eight locales, over 1.2 billion people who might one day access the Web can do so through a localized version of Firefox. Staggering!!

Of course, it’s not fair to use this metric because a much smaller percentage of that great number has access to the Internet.  Luckily, Gerv constructed a model that he blogged about some weeks back that I found helpful in determining percentage of the Internet population covered by Firefox 3.  We can now see the impact of adding new languages to Firefox 3 through Gerv’s model.  Thanks, Gerv.  With the addition of these localizations, Firefox 3’s coverage of the Internet population grew from 88.64% to 93.10%. That is also an amazing jump!

(As I mentioned above, this is based on a rough model constructed by Gerv that we presented at the Mozilla Summit.  We make a lot of assumptions and the data is incomplete, but Gerv has explained his methodolgy in his post.  I’m looking to improve this model going forward so we can be even more accurate.  It’s close, but not perfect.  Ping me if you’re interestd in helping.)

Regardless of the numbers, this was a heroic effort by the community of localizers and by the l10n-drivers team.  You’ll find me passing out congratulations when it’s deserved, so I would like to acknowlege the following people for this effort:

L10n-drivers team:

• Axel Hecht (Pike on IRC) — he probably won’t tell you, but he didn’t sleep much at all over the past couple weeks, working very hard to make this happen.  Amazing work by Pike.
• Mic Berman — Steadily guided localizers through the Web services components of these browsers
• Pascal Chevrel — Great work on managing the web parts of the localization process
• Wil Clouser — Always helping with web dev and clarifying bugs when possible
• Chris Hofmann — Answer questions, keeping up enthusiasm, and developing the community of localizers
• Tim Riley and the QA team — Testing away to perfection…

Localizers “and team”:

• Bengali:  Runa B and the team
• Hindi:  Rajesh and team
• Kannada:  Shankar and team
• Marathi:  Sandeep and gang
• Telugu:  Krishna and the team
• Galician:  Javier et al.
• Icelandic:  Kristján and team
• Thai:  Isriya, Patipat, Arthit and everyone

It’s probably goes without saying, but for me, it’s hard to describe the amount of work and coordination that goes into this process.  Everyone has put in such a great effort to make this happen.  If you’re out having a pint of your beverage of choice, please offer a toast to these people because they deserve it.  Congrats to everyone!

On Monday evening (Pacific time), Pascal filed bug 451128, requesting that our localizers translate strings for the major update “billboard” that we push to end-users for the major update from Firefox 2 to Firefox 3.

Today, Pascal resolved that bug as fixed.  In four short days, our community translated the strings for 44 different locales.  Pascal’s comment in the thread is a nice celebration of the hard work that went into this effort:

“With this last translation, we have completed the task for 44 locales in 4 days! Congrats to all of you…”

Perhaps that’s not enough emphasis or thanks, so let me also chime in with a huge thank you to the localizers and to Pascal for managing the process!

How to improve?

With this acknowlegment, it also seems best to mention some areas where Mozilla can improve to make the process better for localizers so that next time it’s not such a crunch.  I thought of two areas specifically related to this last effort:

• Timing:  Although our community completed this in four days, it would be nice if we can give our localizers the work to be translated with much more time in advance.
• Details:  We should hammer out small details (like iframe window sizes so teams translating know what they have to work with when translating) and communicate that early.

Did I miss other points?

One of our team’s quarterly goals is to draft an l10n requirements document that we will pass over to the development team.  Bits of information like the above will go into that document.  As we get closer to writing that document, I’ll be sure to call for ideas.  How can we make the process better as it relates to what we can communicate to our development team?  Your comments are welcome.

But, just to reiterate Pascal’s point, thank you to all the localizers.   What a terrific effort!

Take a look.

This is a list of localizations in our Mozilla ecosystem, with contributors and who “owns” what on each of the teams.  I spent some time yesterday updating this page (same as above) yesterday.

If you are a localizer, feel free to take a look and make edits where necessary.  We should make sure this is as up-to-date as possible, so if you think there should be a change, please do so.

This post is meant mainly for localizers, so I will cross-post it on the dev-l10n mailing list.  But, I thought I would share it with Planet Mozilla.

This is more of our effort to improve communication with localizers and to eliminate the number of outstanding bugs that have not had activity or are waiting for a response from someone on the l10n-drivers team at Mozilla Corporation.  We decided to resolve a number of bugs as “incomplete” where individuals had not responded to pings for several months (and in some cases, over one year).

I inserted the following note in each bug to the localizers:

“Hello:  Due to the lack of recent activity on this bug, we are going to resolve it as “incomplete”.  If you choose to restart translation activity, then no problem at all.  Please notify the l10n-drivers team using the l10n mailing list (dev-l10n@lists.mozilla.org) and file a new bug in Bugzilla.  That bug should indicate that you are interested in translating the upcoming version of Firefox, which is 3.1 right now.  All the best, SethB”

Here is a list of the lanaguages and bugs I committed as incomplete:

301564 [ig-NG] Igbo: Firefox, Thunderbird
309575 [ky-KG] Kyrgyz: Thunderbird
370204 [gsc] Aranese: Firefox
398393 [ti] Tigrigna: Firefox
407543 [ht] Haitian creole: Firefox
412994 [kk] Kazakh: Firefox

Over the next few days, I will be working on pinging many individuals who have filed registration bugs, but have had recent activity.  Most likely, we will ask these individuals to participate in the upcoming localization of Firefox 3.1.  Obviously, a lot has to be done before any of the localizers need to worry about translating the next major release of the browser.  (for one, we need to get builds and tinderboxen set up…)  But, we intend to start the communication process with the new teams who seem interested in participating in the upcoming release.

I found this summary to be very valuable because Jesper asks a lot of great questions about each tool.  He then formatted the responses to allow for a visual comparison of the tools.  The document has columns filled out with the following information:

• Name
• Description of the tool
• Author(s)
• User Interface
• Used by
• Requires install?
• Requires account?
• Pull / checkout
• Push / checkin
• Glossary
• Translation memory
• Show missing strings
• Automated testings
• Calculate % done
• File formats
• Preserve file layout
• Create language pack
• Links / more info

I’m writing about this because if you know of other tools or use other tools, you might want to contact Jesper or add it to this list yourself.  Jesper posted this to the localization mailing list, so hopefully those who are not a part of that list can contribute if they know of other tools used by Mozilla localizers.

Eventually, we will need to add Verbatim, which is being developed by Wil Clouser and others in the Mozilla Community.  Also to be added one day soon is Silme, which is being developed by Zbigniew Braniecki (Gandalf on IRC) and Adrian Kalla.

If you have any questions, please ask.  For instance some of the terms above may not be familiar to you.  If so, just ask, and we’ll respond in the comments of this post.