Glade-3 in packs?

Danilo Šegan danilo at canonical.com
Wed Nov 28 10:10:10 GMT 2007


Hi Djihed,

Today at 10:03, Djihed Afifi wrote:

>> It's not our decision: it's a technically hard problem.  The only
>> assumption we make is that packages will contain proper POT and PO
>> files.  If that doesn't happen, it's easier to fix it in a package,
>> than to try to construct a reasonable POT from all the PO files
>> (i.e. you may end up with too many messages for translation which are
>> obsolete, thus making translators do unneeded work, you may have
>> conflicting messages in different PO files, etc).
>
> What's technically hard about compiling existing .po files from
> upstream?
> You don't need the original .pot files for that.

That's something Ubuntu packagers have to deal with, if they want that.
I can imagine why they may not want to do it, though (i.e. special
casing for a handful of packages: it's ugly, not hard; it's easier to
fix those few packages which have the problem by hand; etc).  FWIW,
that's what is currently being done for 'universe' packages.

You may talk to Martin Pitt (CCed) about what he thinks of it
(i.e. maybe not run 'strip-translations' when no .pot file can be
found in the tarball).  However, that will make it harder for us to
track down problems like this (though, this is easily solveable: just
report a bug for the package once this is done), and I still think
it's more valuable to provide updated language packs.

What I was talking about a technically hard problem is constructing a
POT file when one doesn't exist.  As far as LP is concerned, that's
exactly the case, and we require POT file to ensure at least some
level of correctness.  We have an idea how to solve that, which would
be equivalent to just "compiling existing .po files", but we don't
like it for the reasons I mentioned.

>> No, but I probably miss your point.  What they, however, do lose is
>> the ability to get updates with future language pack updates (the plan
>> for Gutsy is to issue them monthly). 
>
> I meant - Since Launchpad and the language packs can't get the glade-3
> catalog* - or any other package which can't build them, just roll the
> translations with the package itself. 

Yeah, that's doable, but would have to be done on the Ubuntu side of
things.  I don't know much about it, but I've CCed Martin.

>>  Ubuntu is so far the best
>> distribution when it comes to propagating translator work to their
>> users (imo, though there's a lot to improve still), and that's solely
>> due to Launchpad/Ubuntu
>
> No, it isn't. It's not better than a distribution that just rolls
> everything with their original packages, and does no language packs.

Have you never been late to provide updated GNOME translations by a
few days (i.e. a maintainer rolls out a tarball 3 days before the
deadline, and you do the update after that)?  I am pretty sure you
were, and if you want your translations in your distribution, you need
to wait for the next major GNOME update in it.  With Ubuntu, you just
go to Launchpad, upload your translations, and wait for the language
pack.

> Sure that option may have drawbacks like bandwidth, but omitting whole
> existing translations is not one of them.
>
> <I'm an ubuntu user btw>
>
>> 
>> > Not the right decision I believe.
>> 
>> It's technically very hard to do anything else properly.
>
> Again, what's technically hard about rolling translations with their
> original package?

We can solve this either on the LP level, or on the Ubuntu level.
I was talking from LP point-of-view.  From Ubuntu POV, it's not up to
me to say how difficult it would be, since I am not going to be the
one doing the work :)

>> Anyway, this is not because of Launchpad, this is because we're
>> providing language packs for Ubuntu, and want to make sure that some
>> data is at least correct.  Launchpad itself works for much more than
>> just intltool-enabled modules, so it can't regenerate POT files in the
>> way damned-lies (on l10n.gnome.org) does.
>
> Either way, it is because of the bureuacracy in the middle. Bureaucracy
> can't handle them? just roll it the old fashioned way!
>
> (I'm repeating myself here)

This may require more effort than just fixing a package, though.

>> And, fwiw, this might as well be broken behaviour in upstream Glade3:
>> i.e. standard intltool build rules always create a POT file and put it
>> in a tarball.  There's a lot of sense in that, and not least of them:
>> how would you start translating a package if it comes with no POT?
>
> If I am an end-user, I don't want to translate a package, my priority is
> to get it. If I decide to translate a package, well in the case of
> Glade-3, I do as I did and go spend hours on the upstream package to
> translate it.

And we want to make Launchpad a way to find faster how to contribute
to translations anywhere.  Glade-3 is a special case for you, since
you are Arabic GNOME translation team leader, and Glade-3 is
translated in GNOME Subversion.  But there are other use-cases as
well (like, you are Arabic translator and run into software which you
don't know where to go and translate).  I am not saying Launchpad is
solving them or solving them perfectly yet, but that's where we want
to be.

> And then sit and wait for ubuntu to deliver them and wonder where my
> work has gone.

Reporting a bug was the right thing to do.  Bugs happen.  How many
other translations have you not seen in Ubuntu because of this?  If
this is the first time you've hit this bug, I'd say it's not a big
problem (and I know you've been translating GNOME for a few years
already, and you've contributed a lot of translations).  It might be a
problem that nobody caught this during Ubuntu pre-release testing, but
I guess there are not enough Ubuntu Arabic translators testing Ubuntu
(and Ubuntu mostly a community-driven distribution).

We may disagree on the value of the Ubuntu/Launchpad approach to
translations, but I am pretty confident this one is not such a big
issue: only few packages have the problem, and they are easily fixed.

Cheers,
Danilo



More information about the ubuntu-translators mailing list