[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

My suggestions for i18n

On our web-pages we state "Debian goes International" but still only a few
web-pages are available in non-english. In my opinion we have to start
translating ASAP because it is an awful lot of work which can't
possibly be tackled in one release-period. Let's aim for "Debian 4.0
Hurd/GNU/Linux i386/m68k/alpha/sparc/... International edition" :-)

What needs to be translated:

  1. package descriptions
  2. installation scripts
  3. program messages
  4. program documentation

The following suggestions will mostly deal with the translation of package
descriptions, the multilingual installation scripts are deeply connected
to the auto-installation/message-catalogue problem which is already in
discussion. Ideally the upstream-authors care for translation of their
programs. Realistically, they don't, because they can't. This means that a
Debian translation system should be capable of dealing with .po-files,
too. 4. is a hot issue because it is directly connected to the "free
documentation"-stuff. Our only chance to get a real multilingual
distribution is to have free docs, meaning that we get the sources (no
.ps!) and are allowed to distribute translations. But that's a totally
different subject...

First of all, who should translate? This can be done by

Translation teams
A translation team is a group of persons all dealing with translations to
a given language. Why do we need those teams? The point is that
translation seems to be very easy: You read a text in a foreign language
and you understand it. The next thought is: "When I understand this stuff,
of course I can write it down in my own language." Usually the problems
start with translations of standard phrases in the source-language which
don't exist in the target-language. If we want to get reasonably good
translations we need at least a simple database of common translations for
each target-language (easiest solution: Tab-delimited wordlist), because
nothing is more confusing for the reader that different translations of
the same terms.
The next important thing is that we need non-programmers in the teams.
Personally, I can express myself better with in a programming-language
that in inter-human-languages and I heard that this is quite common for
programmers (hopefully noone of you is offended now). This suggests that
a translations system must be usable for mortals :-)
Annother useful thing for translation teams could be some sort of   
job-system, a kind of priority-queue of texts to be translated. When for 
example a package changes the description the translations should be
adapted ASAP. It could be something like a priority-sorted HTML-list to
check out the texts. At least something that can be used by normal 
If someone knows how GNUs translation teams are working, please tell 
me/us. The ABOUT-NLS just mentions that those groups exist and I would be 
very interested to know how they are working.

How to include multilingual descriptions
The main problem is that we have a working scheme which has to be changed
to allow descriptions in other languages. Those are the solutions which
came to my mind (missing: the Perfect Solution (tm)):

1. Use one Packages.gz for each language and set a symbolic link
   Packages.de.gz -> Packages.gz. This would only require minor changes
   but is clearly in the category "quick hack". Also the changing of
   symbolic links on CDs is rather difficult. And at least all routines
   dealing with the download of "Packages.gz" need to be changed.

2. Extension of the description field. We allow something like 
     Package: foo
     Description: ... [this is implicitly english] ...
     Description[de]: ...
   for all different languages. This requires changes to dpkg, dselect and
   apt and will blow up Packages.gz

3. Something in between: We use the normal Packages.gz and additional
   files with only descriptions in one language in it, 
   (descriptions.de.gz ?). The entries would be something like:'

   Package: foo
   Description: ...

   Package: bar
   Description: ...

   The programs using the descriptions would have to check if there is a
   translation in the wanted description-file and use the one in 
   Packages.gz as a fallback. This solution would require heavy changes in
   our programs but would allow just to download the needed languages.

Sending translations to Debian
Our translations system could be accessible by mail. Whenever someone has
finished a translation she/he could send it to translation@debian.org. The
incoming mails are PGP/GPG-signed and are checked if they are sent by a
maintainer/member of a translation team. The format of the mails could be
like this:

Package: foo
Description: ...

We should also have a possibility to send .po-files to translate all

Package: bar
Version: 0.07
po-file: <end-tag>

Ok, that were my thoughts regarding translating, please correct me, please
don't be offended, ...


Thimo Neubauer <thimo@debian.org>
Debian 2.0 released! See http://www.debian.org/ for details.

Attachment: pgpb4aF64BMYp.pgp
Description: PGP signature

Reply to: