Dienstag, 10. Juli 2012

Some questions on l10n and my answers


Some time ago Jeff Beatty asked localizers to fill out a survey about motivations/reasons for doing localization of Firefox.The results of the survey can be read in Jeff's blog Why we localize Firefox. The survey asked for additional comments and I raised my hand. Jeff then sent me a barrage of questions, which I am copying below, with my answers to them.

Questions and answers

Jeff:  you wrote that  "from initial go-get-em drive it has turned into watch-for-deadlines routine." Do you attribute the evolution of a "go-get-em drive" to a "watch-for-deadlines routine" to being a natural result of contributing to the project over a long period of time? Or to any of the current l10n policies, communications, or procedures?

Vito: what I meant by  "Go-get-em drive" was getting into the release process in the first place. Firefox in Slovenian language has been around since ages, but it is a different story with Thunderbird and Fennec. With Thunderbird it took us a whole year, if not more, to get fully involved in the release process. With Fennec it is a similar story, with somewhat changed border conditions: after a year of active involvement we have nightly builds, but as yet no official release.

As you see, I am not limiting the context just to Firefox, as was your initial intention.

The evolution from "Go-get-em drive" to "watch-for-deadlines routine" is thus (of course) a natural result of contributing to the project over a longer period of time. It just has to be understood as a sign of persistence. Note that  77 out of 106 languages represented in l10n repository can boast with an official 13.0.1 Firefox version. I wish the remaining quarter all the best on their way to the "watch-for-deadlines routine".

Instead of "watch-for-deadlines routine" one could talk about  "riding on the release bus", were it not for the switch to tighter release cycles. Fore me it resulted in kind of a tunnel-vision response, with the dashboard record for Firefox, Fennec, Thunderbird and Calendar (times Aurora, Beta and Central) on the one side and the ticking deadlines bomb on the other side.

As regards your second question: I think the major reason for "watch-for-deadlines routine" is the time squeeze. I am sure  there's been a lot of soul searching and fighting over possible fall-outs involved in the decision.This is just one more negative side effect that was to be expected.

Jeff: others, I'm sure, have shared this aspect of your experience. Do you feel that there is anything we can do to help re-inspire a "go-get-em drive" in localizers?

Vito: what helps most, is attention and a positive feed-back. That applies above all  to the language teams among the 25% that are not yet fully there. I wonder if anybody, starting on this journey knows, what it takes (as of end of June, the aurora SL repository contains close to 70.000 words). So what one needs is endurance first of all and any help to sustain it. Note, that there's differences between different language teams, from a lonely wolf to a well-oiled localization machine, involving several Mozillians (see my blog on localizing SuMo)  so it is not one-shoe-fits-them-all.

Jeff: Many communities struggle with remembering deadlines, causing their localized releases to fall behind with each release. What suggestions would you give to remind communities about upcoming deadlines without giving the impression of being driven by deadlines?

Vito: with three channels (central, aurora, beta), each involving two product repositories, mozilla and comm,  and one l10n repository, it is pretty hard to keep your ducks lined up properly - they get out of of line  every six weeks anyhow. Anyhow, I have a strange feeling, that is one, two or even three repositories too many.

One more point: there's absolutely no need for localizers to have complete product repositories on hand. The size of my local  Aurora repository is 1.4GB, out of which I need just 5.44MB in size. I really have no need for those extra 1394.56MB, the en-US tree, extracted from  up-to-date product repositories, is all I would need. I asked for it already several times already:

Hint 1: provide the en-US tree in l10n/releases,  to serve as a source and template for a given channel. This can be done with one single compare, run once and for all languages. If this is not doable, then whoever has an en-US tree on hand, drop me a note.

Note: with this hint I am probably risking  a barrage of Why-dont-you, Havent-you-read and  I-told-you-so comments. My comment to all in advance:"where were you all, when I needed you".

As regards the question about helping localizers keep deadlines ... maybe a polite hint on mozilla.dev.l10n, like "seven more days to ..." with the URL to more about it, would not look too invasive.
    Jeff: You're also quoted as saying, "it is NOT just Firefox we are talking about (or at least localizing). For instance Firefox alone without SuMo is a job halfway done. A little broader perspective would be welcome." Do you feel that Mozilla l10n is too centered on Firefox desktop? Please explain.

    Vito: I just said that when it comes to localizing, there's more to Mozilla than just Firefox. This pizza has several slices:  products, MarketPlace, Persona, Mozilla Support (Firefox, Thunderbird, Mobile, Desktop), Corporate Web presence,Verbatim projects and so on. A localizer thus has to communicate with a number of pizza slice owners, who all in their own turn compete for the localizers' attention and favor. Think of it like this: ab_CD team owns the ab_CD localization market and the Mozillians on the other side of the fence are in competition with each other for its services.

    Eventually one can not be everything to everybody - that's where either the specialization starts, assuming there's a team in place.If there's "more than one person, but less than two", which is often the case, the game is then played by the ear. And sometimes it's the loudest, and not necessarily the overall optimal alternative that wins. Maybe this Brownian motion is the best one can expect in a given situation - provided there's a focusing discussion or a MozCamp now and then.

    Jeff: To provide you with some perspective, I chose to ask about localizing Firefox since most l10n contributors are attracted to the project because they have been loyal Firefox users. Do you feel that localizers on other l10n projects are frequently overlooked by the l10n drivers?

    Vito: I guess drivers are in the same unenviable position as we are: they have to decide what to focus on and what to let go or simply ignore. Instead of "frequently overlooked" I would thus rather say "practically invisible": other l10n projects are out of the l10n drivers' attention span and the localizers' involvement elsewhere, even if it is the one and the same team, i.e. same persons,  plays no role.

    PS: have a comment? Diverging opinion? Hint? Experience to share? Write it down here!

      Sonntag, 22. April 2012

      Firefox SL - a look into the rearview mirror

      Here the last half a year or so of data, culled from the biweekly l10 metrics reports.

      What I see in these diagrams

      • Blocklist FX - SL locale: solid long-term level, with the Christmas and  New Year knick and possibly a slight downtrend lately.
      • Blocklists - past Fx versions: getting users move to Version 8 was a hard battle. Judge yourself, if it was won or lost. The version mix exhibits a nice, regular pattern since.
      • Downloads -  past Fx versions:the reluctance to move to the 6-week cycle is pretty much evident here. After the initial hype with version 8  the number of downloads has dropped drastically (due to addon and plugin problems I would say). The total number of downloads per version, however, seems to be on the rise.
      In any case, we need back those users on the left of the downloads diagram. Maybe they are waiting for the Mobile Fx in SL locale. It's coming in version 12,

      Mittwoch, 4. April 2012

      Mr Firefox, I presume

      It is in a way funny, how long it has taken me to face this question. It is rather simple. The foxes need two genders to survive, procreate and flurish - and this applies to Firefoxes too. Our Firefox, is it a male or a female?

      For French (le renard), German (der Fuchs), Spanish (el Zorro) is a male. On the other hand for Italians (la volpe), Portuguese (la raposa - not sure) and Slavic languages (e.g. Russian, Bulgarian, Croatian, Serbian, Macedonian and my mother language Slovenian) fox is lisica, i.e. it is neither it, nor he. It is a she, i.e. female.

      So there's a non.negligible group of languages where the gender difference is still present and where the female alternative (for a change) prevails and where foxes, Firefox included, are ... ladies.

      Here's now the text I had to localize for Firefox flicks:

      You're impossibly fast and strong. Your fur is soft and warm... and bright orange.Your eyes are kind and you're always thinking of others.You're never closed and you never compromise.

      ... and then comes the scene, where the girl says to the man off-screen - and I kept typing -

      Please don't make me choose. Because it'll be her.

      Oops, the English source says him! I quickly corrected "her" to "him" in the translation and then said to myself: what a nice, surprising gotcha. What a double-entendre... Then I had a laugh, drew a deep breath and repeated: Firefox is a man, is a man, is a man. Firefox is a man!

      In Slovenian, and probably in other Slavic languages, a "he-fox" is "lisjak", so theres a way out of the bind. Except that, at least in my mother language, "lisjak" has a questionable sound to it, like cunning, double-dealing, Artful-Dodgerish.

      Remember Josephine in Some like it hot, repeating "I'm a man, I'm a man..."? Pretty similar situation: Firefox is a man, is a man, (sigh) is a man!

      But then, nobody is perfect.

      Montag, 12. März 2012

      You're telling me you do L10n? Show me your commits

      The reason for this blog is the following table from a corner in the Mozilla land. The name, email etc columns are uninteresting. Have a look, however, at the Role column: just one person out of 28 has not   raised his hand for translation and localization. Means that more or less everybody in the list is an active localizer?

      The localization at Mozilla can mean different things:

      Looking just at the commits for Mozilla aurora versions and into the verbatim statistics gives us a  picture, that's radically different from 27 out of 28 above. Only 8 persons out of 28 are active committers. A similar situation in Verbatim: a very lenient selection criterium  (5% and more of the work done) eliminates 17  entries from the list of 28.

      One must conclude that the entry "l10n" in the role column does not necessarily mean active involvement. The best one can hope is that it means a raised hand. In the busses,bringing the pilots to their planes during the Korea war, the following rule was enforced: standing-room-only for anybody, who had yet to shoot down an enemy plane.There's been a lot of people standing around in our bus. Time to give them an opportunity, to challenge them, to take a seat.

      Here's some hints on how to check the localization status for your locale:
      • L10n product commits: can be deduced from http://hg.mozilla.org/releases/l10n/mozilla-aurora/locale (for instance on  http://hg.mozilla.org/releases/l10n/mozilla-aurora/el for the Greek community).
      • Verbatim activity: you get the stats for Verbatim activity in a similar fashion from http:/localize.mozilla.org/locale, for instance on the page https://localize.mozilla.org/sl for Slovenian.
      • SuMo stats:  are a little more complicated to get: you may want to have a look at the status of the site translation  first - translating support articles without localizing the site first does not make much sense. The status of the Support knowledge base and its localization can be deduced from https://support.mozilla.org/locale/localization page (here's the Croatian page).
      • Mozilla.org: last but not least, the Mozilla web sites. You can start at the top node on locales .

      Montag, 20. Februar 2012

      SuMo - a look at locales

      The Support for Firefox site is proudly multilingual. 71 languages, from Akan to Zulu can be selected, which is rather close to the number of localized Firefox versions  (85 and counting). What does it take to localize SuMo?
      • SuMo Site: The localization runs within the Verbatim enviroment. The size of the job is close to 6000 words in size. 
      • Articles: together with the templates used there's about 300 units to be localized. In the OmegaT project, used for the SL locale, the 93 articles translated amount to ~54.000 words. Ballparking one single article then at 600 words, the complete phase space of KB articles must be close to 180.000 (!)  words. To stay within reasonable limit, the top 20 articles should suffice, but that still means about 12.000 words.
      The amount of text to be localized to a reasonably complete degree is thus 18.000 words. It does not match the total size of Firefox localization (ca. 180.000 words),  however, these additional 10% however can make quite a difference. Think of a Firefox user with a problem: when an English version of the help pops up on pressing F1,  the fun is gone for all non-speakers of English.

      Looking at the status

      Every language, recognized by SuMo, has a dashboard, showing the current status of the localization - for instance for Slovenian language (a little plugging for one's own cause does not hurt;). There one can see, how many of the top 20 articles and so-called templates are up-to-date in the selected locale and how it stands with the localization of the SuMo site itself.  Current status for SL is all green (top 20 articles and templates localized). 23% of all the articles has been localized and as regards the SuMo site itself,  it is fully localized in SL (evident from the Verbatim dashboard for SL ).

      That much about SL - what about the other 70 languages? The data for the diagram below have been collected from the Dashboard pages:

      The diagram shows the number of localized articles and templates versa the number of localized top 20 articles. The four clusters (from left to right) can be identified as follows:

      • the spirit indeed is willing, but the flesh is weak:  this is by far the largest group with 71 language groups present. They are at the beginning of the SuMo journey. 23 of the languages in this group  have not one single text unit translated. This is one third of all the languages advertized in the drop down list box. Altogether 37 languages (i.e. a good half of the whole portfolio) are in this group. One could just as well tag it as "Huh!?" group.
      • moving up: there's three languages on the move from the start-up group to higher grounds. we wish them all the best on the way.
      • room at the top : eight locales have it made (the top group around 250 articles+templates and top 20 from 15 on). Let us mention them by their names: Dutch, Japanese, Polish, Italian, French, Spanish - Castillian, German, Chinese (PRC)
      • let's be practical: the two languages with top 20 in green, but with a significantly lower number of translated articles (bottom right), are Czech and Slovenian. I can not speak for CZ, but in the SL case it is a question of diminishing returns. Practical example: for SL there was one single Helpfulness vote the last 6 months for #20 - Importing favourites and other data from Internet Explorer.
      ...Two months later

      The four (or three and a half) groups above can be identified elsewhere as well - for instance in the  the completness of  product localizations as well as their portfolio. The locales in the "room at the top" group are no surprise. What may be surprising at first glance, is their absence from SuMo fori (with the laudable exception of FR), but just at the first glance: they have adressed the question of support some time ago with the help of their own fori etc. I have yet to find out how they make the localized material in Knowledge Base to their language group - I assume they do, but have bo proof for the assumption.

      PS: this is just some of the facts I have culled from the Verbatim site. See also Thar she blows ...