On Sat, Mar 27, 2004 at 02:18:29PM +0100, Denis Barbier wrote: > Jure Cuhalev wrote: > [...] > > Once done translating po message catalog we have to return it back to > > Debian. There are again a few ways to do this: > > - post a wishlist bug report against the package > > - commit it to CVS/SVN repository > > - mail it to mailing list or to the person maintaining it > > > > While the first method seems to be preferred there are many packages > > that need to be translated and reporting bugs against each and > > everyone > > is not something that translator wants to do. Therefor I propose and > > improvement in this field. The thing I miss in Debian is central point > > where translation team coordinator can commit translations of his team > > and get new, fresh ones - without having to deal with BTS for each > > package template that gets translated in the process. > [...] > > There is one problem you have to address: you have no control on what > is packaged, so you cannot be sure that the translated file comes from > your central repository. For instance the maintainer might decide to > translate it himself, or he receives a translation through the BTS or > via direct mail. If a maintainer wants to use this framework, he must > only accept translations from it, which is a constraint for translators. > This framework has to be carefully designed so that this drawback is > outweighed by its advantages. In my mind the system would have been a repository for translations, nothing more. That could be useful so that I can put in a central location the translations I have underway. It could be possible to make tools for translators allowing them to get the material to translate with a command-line tool (instead of wget on your pages). Nothing revolutionary, but anyway. The second step would be to have a tool for packagers to fetch available translations while building the package. If they get translations from another source, the other material would get automatically merged to the one in the repository during the transmon-check run. Then, if we ensure security on this system (request translators to PGP sign their work with recognized keys), we can design an override system merging the work of translators to the package when they are uploaded by the packager. Another step (even more far away from now) would be to have a reliable way to build automatically "l10n upload only" version of the packages, and have a mecanism building and uploading them automatically when translators commit their work on the cvs/svn. Well, nothing of this is new, I already spoke about this dream several time, but I just wanted to emphasis that even if the cvs/svn repository is not very useful on its own, it consitutes a crucial part of a reliable and efficient l10n infrastructure adapted to a decentralized organization such as Debian. Mmm. When I read again the last paragraph, I get the feeling that I should stop writting empty blurbs to persuade people to get my application for researching position >:=D Bye, Mt. -- Source is provided to this software because we believe users have the right to know exactly what a program is going to do before they run it.
Attachment:
signature.asc
Description: Digital signature