On translation regressions due to freedesktop.org dependencies

[Christian Rose <menthos gnome org>]

There have unfortunately been significant regressions now in translation
coverage as we are in the phase of moving over to using freedesktop.org
technologies inside GNOME, like recently with the MIME system, the MIME
types and their translatable names and descriptions.

Unfortunately, there's no short-time solution to this in sight as far as
I can see, and it looks like this will be the case not only in the short
GNOME 2.8 time frame, but beyond that aswell. The problem is mostly an
administrative one, and not a minor one mind you, which is why it looks
difficult to resolve.

This is basically the problem:

While some key developers have access to freedesktop.org CVS in order to
be able to maintain their software, many other contributors have not.
This includes translators.


Vincent van Adrighem was recently so kind as to research how this
affected the MIME translations¹. They are easily comparable since many
of the fd.o shared-mime-info translations are identical or very similar
to the gnome.org gnome-mime-data ones.

The results were devestating. For gnome-mime-data, we have translations
into 67 languages, many of them 100% complete.
The current status in freedesktop.org is only 13 languages, and most of
them not even 50% complete.

As there has been changes in shared-mime-info compared to
gnome-mime-data, simply copying over the translations unfortunately
wouldn't solve the completeness problem. The only real solution is for
translators to have access to fd.o.


"-- Well, then why don't you apply for accounts?" one might rightfully
ask. Well, we did, we still do, and it looks like we will still keep on
doing it. Out of all GNOME translators, I'm currently only aware of one
(1) GNOME translator who has yet managed to get access to fd.o, and it
seems he got it by sheer luck.

"-- Well, that's a start, isn't it?" Possibly, but translation is labour
intense work, and one could employ a single person full-time day and
night and he still wouldn't be able to commit all translation updates
that is done in GNOME cvs on a single day. We have at least 88
translators who have access to GNOME CVS (and the figure is probably
still a bit higher, since all translator accounts are not marked as
such, 120 would be a more reasonable guess IMO). So the problem is a
quite big administrative one, and it seems currently the administrative
process isn't even there at fd.o, and noone to the best of my knowledge
working at resolving it.

"-- What about the Translation Project then?" Right, in the past we
suggested fd.o maintainers use the external Translation Project², which
is a project that allows maintainers and translators to cooperate
without a common CVS.
However, as this is done by synching and incorporating pot and po files
back and forth by mail, which can be very time consuming and also easy
to forget to do, many fd.o maintainers have been reluctant to use it on
a frequent basis. This, as in the case of shared-mime-info, leaves
translators with only a more than half a year old snapshot to work with.
Suffice to say, much can happen in half a year, and certainly so with
shared-mime-info. I can very well understand maintainers here -- it's
not convenient by any means if you compare it to the GNOME process where
translators work essentially by themselves directly in CVS, and if I
were a maintainer I would much rather code than having to do the
synching administrativa required with the TP.


So where does this bring us? Obviously there are a tremendous amount of
benefits in the cooperation done at freedesktop.org. There's no question
about that, and doing anything else would be silly.

However, in the long term, we should probably try to consider in advance
how adopting particular freedesktop.org pieces of software, and
incorporating such dependencies, into GNOME affects other established
qualities we might have, like current translations and documentation
etc, in order to not introduce bad regressions.

In the short term, I think the battle is lost for GNOME 2.8. Even if we
had 88 fd.o translator CVS accounts ready tomorrow, there's too little
time left to the GNOME 2.8 release to expect any comparable translation
coverage like we had before. While some translators work day and night, 
usually need a lot more time to get some spare time aside to do some
translation work, so there is often a significant delay in the process. 

So the sad news is that while we could brag that we had at least 80%
coverage into 36 languages (and a dozen of thosefully complete) in GNOME
2.6, we will probably not end up with more than a dozen 80% complete
translations in GNOME 2.8, and perhaps not even a single fully complete
one at that.
That is, unless we suddenly declare the (very) visible MIME info and
descriptions not part of the GNOME desktop and not our problem(tm). But
that would be unfair to users, and still, for all what it's worth, a bad
regression.


Christian



¹ http://mail.gnome.org/archives/gnome-i18n/2004-July/msg00115.html
² http://www.iro.umontreal.ca/contrib/po/HTML/