Re: Ubuntu vs. Debian translations
On Tue, Jan 23, 2007 at 09:47:18AM +0100, Danilo Šegan wrote:
> Today at 6:37, Peter Mann wrote:
> > every package should use any primary system for translation: Rosetta or
> > Pootle or svn/cvs or only http po download + wishlist bug report -
> > depending on author choice ... but every packager/maintainer and
> > translator should know about this translation system ... so every
> > translation is updated in upstream ...
> That's not entirely possible. Distributions *do* introduce changes
> (patches such as panel menu changes in Ubuntu), and any translation
> system used by such modified software should *not* automatically push
> translations to originating software as well.
yes, but common parts of translations can be translated as "upstream"
part ... and small differences can be maintained by specific
distribution changes ... now we must translate common parts a few times,
because of bad cooperation between all of us (author, maintainer,
> > and there is possibility to include distribution specific strings in
> > upstream - like e.g. Ubuntu related strings in debian-installer big po
> > file, which are never used in Debian, but strings are ready for Ubuntu
> > (and one translator can translate strings only once)
> The idea of minimizing work for translators is present everywhere.
> Debian probably wouldn't want to include Ubuntu-only strings in their
> translation files.
for debian specific things yes - because Debian is for those things
upstream - debian-installer is real example ... i translated Ubuntu
strings in d-i big po file
> > yes, i know about exporting po files, i used export sometime - but why
> > is so hard to provide clean http access to po files (only for download) ???
> > are po files secret???
> Of course they are not secret.
i know, this was my sarcasm ;-)
> This is simply for the performance sake.
> We have some plans to improve performance, when we may also enable
> direct PO access. However, since we're sometimes talking about large
> PO files, it may still mean timeouts when exporting (our current
> web-backend processes are limited to 30sec runtime).
ok ... but IMHO this is wrong direction: from GUI to backend ...
i prefer good backend (e.g. svn, bzr or something similar) and then
additional GUI ...
> > i mean if maintainer can import all translations from Rosetta to Debian
> > ... so untranslated Debian strings could be translated after import -
> > and maybe some fuzzy strings remain ... but better solution is
> > cooperation between upstream and maintainer and SOME translation backend
> > system ...
> I agree it's better, but that's impossible to achieve in practice.
that was idea only about single one-time step
> Translators using Rosetta are not always the same translators working
> in other projects. This means that such automatic "cooperation" would
> be a source for conflicts, arguments and bad press.
i don't prefer automatic cooperation ... every upstream author
+ packager/maintainer should use their own system - and every
translator must know about that system ...
so again ... one string ... one translation ... only once ...
5o Peter.Mann at tuke.sk