[Savannah-hackers-public] Re: Pootle server at Savannah

[Yavor Doganov]

Sylvain Beucler wrote:
> 
> Is there any interest to install it at Savannah?

For the time being -- no, at least as far as web-translators are
concerned.  RMS promised a separate server for our stuff (mostly
because the FSF admins are always lagging, and this would require many
tweaks/interventions, at least in the beginning), although I'd be more
than happy to have such a thing under your control :-)

I know nothing about such web services, and I despise them immensely,
especially for translations.

Among the other tools (references below), I think Pootle is what would
serve us best.  We basically want the following:

* GNUN will update the PO files in www's repo.  We could develop a
  daemon that could speak to the webapp, but it would be better if the
  latter is VCS-aware and can pick up updates directly via CVS.
* The Pootle server should fetch these updated pot/po files
  automatically, and provide a way for the people for whom "using a
  computer" == "using a web browser" to contribute (surprisingly,
  there are many).
* The PO files edited via Pootle should never be installed in www.
  They should be marked somehow in Pootle's interface, and changes
  mailed to the respective co-ordinator (preferrably a diff).  The
  co-ordinator may choose to direct these notifications to a mailing
  list, and only install them *manually* in www after the team's
  review process is done.

Basically, we want to hook the pot/po files in www into some web
interface mostly for statistics (just like Damned Lies --
http://l10n.gnome.org) plus eventually editing for those who don't
know another way, provided that the points above are doable.  Such a
web-service will undoubtedly be a second citizen (i.e. it should not
be required to use it in order to contribute translations), and that's
how it should be.  I don't want to commit a harakiri by doing my
translations in a stupid HTML box :-)

There is a fundamental problem with Pootle, that it chokes on large
data.  That is why the Debian i18n people are still not using their
Pootle server for production work (and debconf templates are short, as
you know very well -- I can only imagine what would happen with the
essays).  I don't remember from the discussions on the Debian i18n and
Pootle lists whether the problem was related to mutliple PO files,
large PO files, or both.

> (or maybe another tool I don't know?)

This is what has been proposed so far; I haven't had the time to
review all the tools extensively (even Pootle):

Pootle     was proposed by the Pootle developers years ago
Entrans    http://entrans.sourceforge.net/demo/main.php
Poliglota  https://tracker.gnulinuxmatters.org/wiki/Poliglota
CLWE       http://www.wiki-translation.com
Ikiwiki    http://ikiwiki.info (not actually proposed by anyone, but
           it is used by the Hurd folks and a native translation
           plugin using po4a is under development)


As for the installation of Pootle at Savannah -- we're not ready to do
anything web-related yet (unless someone with knowledge and skills
chimes in), but I guess having a Pootle installation won't do any
harm.  On the contrary, it would enable the brave (and those
interested in such an interface) to experiment, provided it is not
disruptive for the Savannah hackers' workflow.