> Taking into consideration the list feedback, and some issues that
> came
> up at the meeting, we agreed to accept this substantially modified
> text
> during the blocker review meeting today:
>
> "All critical path actions on release-blocking desktop environments
> should correctly display all sufficiently complete translations
> available for use"
As usual I'm providing feedback only after it has been agreed on. Sorry for that. But I have worked in l10n area for quite some time and would like to add my thoughts. It seems to me that we implemented this criterion somehow backwards.
1. This is not an i18n criterion, this is an l10n criterion. i18n = making the software translatable. l10n = translating the software.
2. I believe l10n criteria are not a good idea.
a) It should be the sole responsibility of the l10n team to provide all important translations in time before release.
b) The missing translations can be easily added anytime post-release as an update.
c) There are a lot of languages, with different team sizes and different people responsiveness. Let's take Czech Gnome translations as an example [1], because it's close to me. The translation is nearly 100%, so it is "sufficiently complete translation". But there are only few people on the team and they might not be responsive in the time you need them. Czech is not a "popular language" from world perspective, but this team size and responsiveness issue might be very well valid even for many wide-spread languages. Will we then delay the release because we will be waiting for a button translation?
d) How do we distinguish an "important language"? It's really tricky. The most common language in the world is Mandarin [2]. Is it an important language? What about Hindi-Urdu, Arabic, Russian or Japanese? All of those have most speakers than let's say German. But do we have more German users or Arabic users? We have no idea. This is gonna bite us every time.
e) Maybe I'm just skeptical, but there is a potential for extra loads of work hitting our shoulders. People could start reporting missing translations into our bugzilla and proposing it as blockers instead of dealing with it in Damned Lies [3], Transifex [4] and other tools. As a translator I would do it ("Cool, they'll postpone Fedora release until our login page translations hit the repo").
3. Even though I don't like l10n issues like Blockers, I believe all of them can be NTH. It should be easy to verify that the new update contains only updated translations and no code changes. In that case there is a minimal chance of breaking stuff. This way l10n teams can easily fix some mishaps even in the freeze.
4. On the other hand, I believe i18n criteria are much more important. It's easy to provide the missing translation in an update (and Anaconda can even download them during the installation process), but it's impossible to translate software when it's not internationalized properly. Then a code fix is needed, which may be more disruptive. Of course we can't force all software creators to internationalize their applications. But I think we pretty much agree that all the most important parts of Fedora ("the critical path") should support localization. There may be software that don't, and the project owners state it's intentional, then it's ok. But for the rest I would be inclined to _consider_ Blocker for high-impact areas (login screen, firstboot, installer, etc). It should always be considered, I don't see a one-size-fits-all solution here.
5. It might be due to my language handicap, but I don't understand the wording of the current criterion. "correctly display all sufficiently complete translations available for use" could be:
a) that national characters don't get screwed up (i.e. white boxes instead of ěščřž)
b) that all translations available in the PO/MO file should also appear in the application
c) that there must be no missing translations (this is the original purpose of the criterion, but I wouldn't have guessed it without reading bug 706756).
My several cents.
[1] http://l10n.gnome.org/teams/cs/
[2] http://en.wikipedia.org/wiki/List_of_languages_by_number_of_native_speakers
[3] http://l10n.gnome.org/
[4] https://fedora.transifex.net
--
test mailing list
test@xxxxxxxxxxxxxxxxxxxxxxx
To unsubscribe:
https://admin.fedoraproject.org/mailman/listinfo/test
