[Freeipa-devel] Reviving FreeIPA translations

Sun May 15 19:34:01 UTC 2016

написане Sun, 15 May 2016 21:51:45 +0300, Martin Kosek <mkosek at redhat.com>:

> On 05/14/2016 01:29 PM, Yuri Chornoivan wrote:
>> написане Sat, 14 May 2016 12:57:13 +0300, Jérôme Fenal  
>> <jfenal at gmail.com>:
>>
>>> 2016-05-13 13:32 GMT+02:00 Martin Kosek <mkosek at redhat.com>:
>>>
>>>> Hello,
>>>>
>>>> As you may or may not know, Tomas Babej left the FreeIPA team as a  
>>>> Red Hat
>>>> employee, which of course does not mean he cannot contribute to the  
>>>> FreeIPA
>>>> project, but that he won't have as much time for contributions as
>>>> previously.
>>>>
>>>> One of Tomas' responsibilities was taking care of FreeIPA translations
>>>> (Translation Maintainer role). As far as I know, there 2 main tasks
>>>> associated
>>>> with the Translation Maintainer role:
>>>>
>>>> 1) Periodically uploading new upstream strings to the FreeIPA  
>>>> translation
>>>> platform of choice, which is the Fedora Zanata instance:
>>>> https://fedora.zanata.org/project/view/freeipa
>>>> The upload should happen periodically, on the right occasions, so  
>>>> that the
>>>> translators (especially the French and Ukrainian translations which  
>>>> have
>>>> 100%
>>>> translated) have sufficient time to translate strings for the next  
>>>> version
>>>> and
>>>> do not have to translate it all in couple days before release. (This  
>>>> was
>>>> one of
>>>> the feedback I heard recently).
>>>>
>>>> 2) Downloading translated strings, Tomas added a short HowTo to our
>>>> Release page:
>>>> http://www.freeipa.org/page/Release#Translations
>>>>
>>>> We will need a new volunteer who would help doing 1) and 2) for the  
>>>> planned
>>>> releases and making sure this process runs. The first task would be
>>>> uploading
>>>> current strings in master as the next release is FreeIPA 4.4 planned  
>>>> for
>>>> June,
>>>> so it may be nice to already upload new strings we have in FreeIPA  
>>>> already
>>>> to
>>>> Zanata, so that they can be translated in sufficient time.
>>>>
>>>> Volunteer(s)?
>>>>
>>>> As part of the learning process, I think it would be useful to do more
>>>> documentation of the steps taken in every translation life-cycle,  
>>>> current
>>>> HowTo
>>>> in Release page is rather vague. I for example did not find  
>>>> information
>>>> how to
>>>> work with translation versions that I saw defined in Zanata  
>>>> (branching may
>>>> work
>>>> similarly as in current FreeIPA git).
>>>
>>>
>>> Thanks to the two volunteers!
>>>
>>> Looking forward to see this happen!
>>>
>>> To reiterate on Martin K. message on uploads, I'd really like to see
>>> regular strings uploads to the master branch in Zanata, say once a  
>>> week or
>>> every two weeks, so that translators can work on smaller strings  
>>> batches.
>>> "Release early, release oftem" :)
>>>
>>> And strings that would be translated twice in a short time span  
>>> wouldn't be
>>> entirely lost because they may stay in the Zanata translation memory  
>>> and
>>> could help translators finalizing dot releases if the specific  
>>> branches in
>>> Zanata.
>>>
>>> And if we can see the upload to master soon, translators can start
>>> working now before the branch for the 4.4 June release.
>>>
>>> Regards,
>>>
>>> J.
>>
>> Hi,
>>
>> Similar thoughts here.
>
> Thanks for feedback!
>
>> Just a note on branches, I think that it is worth to keep the  
>> translation just
>> for the current release because keeping several branches on Zanata (or  
>> any
>> other translation platform) is proved to be not effective from both  
>> sides,
>> developers and translators.
>
> I see. My expectation would be that these branches would be only used if  
> there
> is a bug in the translation, not for active translation. The thing is,  
> strings
> in master may change or may be deleted, so they may no longer be applied  
> to the
> branched FreeIPA x.y.z releases. So practically, we would not be able to  
> update
> the translations for branched release once we branch.
>
> Do you see that expected and acceptable?

Just two examples from the projects that I am involved (three polar  
examples).

KDE (Desktop environment):
1. Developed translation infrastructure (dedicated server,  
specifically-tailored software (Lokalize)).

2. Four translation branches (stable and trunk for Qt4 and Qt5-based  
applications).

3. Automatic message extraction every 24 hours.

4. Inbuild translation integration for releases.

All this needs attention and strict release rules to keep everything in  
sync.

Inkscape (SVG editor):
1. No specific infrastructure.

2. Translation branches are not strict (translators should guess what and  
where they are translating).

3. Manual extraction from time to time.

4. No specific integration or QA.

Medium attention paid from the Inkscape developers.

GnuPG (encryption framework):
1. No specific infrastructure.

2. Strict branches (1.4, 2.0 and 2.1) but no syncing (should be performed  
by translators manually).

3. Manual extraction. No strict release schedule. You are lucky if you  
send your translation in time.

4. Manual integration by Werner when he finds it necessary. ;)

Minimum attention paid from the developer.

I think that FreeIPA in the sense of translation handling should be  
something in between. For not wasting efforts of the developers (it is  
sensible, because of too few more or less complete translations), I  
propose to have just one branch (no switching, no problem of choice), but  
use it strictly when releasing.

It is better to have a translation with several untranslated new messages  
in a branch, than have no update at all because of the release flaws (like  
it is in libvirt now). So it would be a good trade off for me if FreeIPA  
developers promise to integrate translations into each release, but do not  
promise to waste their time for having translation branches. ;)

> It is a matter of just two commands (one for extraction and "zanata
>> push" for
>> pushing the catalog to Zanata). So, personally, I'd like to see the  
>> updates as
>> soon as possible (something close to continuous integration). This  
>> allows us,
>> translators, to react on any glitches in the initial strings and make  
>> the
>> releases perfect.
>
> I think this can be done, there is just the risk that some strings would  
> be
> added during master development and changed later when the code is  
> revisited,
> but I assume this is expected by you - correct?

Yes, that's the common translators risk. But we have an automated  
translation memory for this. ;)

>
>> It would be good if each release preparation process is close to the
>> libguestfs's one:
>>
>> http://libguestfs.org/guestfs-hacking.1.html#making-a-stable-release
>>
>> Just my 2 cents.
>
> Thanks for the tip!
>
> Martin

Best regards,
Yuri