On Fri, Jun 17, 2005 at 10:30:07AM +0300, Eddy Petrisor wrote: > On 6/17/05, Martin Quinson <martin.quinson@loria.fr> wrote: > > > martin> Juste a stupid idea: I know Christian is dreaming of > > > martin> setting up a pootle server on debian machines to ease > > > martin> translation works. What about using pootle for the > > > martin> interface? > > Why more division in the translation effort? Why can't it be _only_ > _ONE_ central place where people can do their work/get info? Because it still has to be implemented. :) > > IMHO, the tricky part will be to interface it with all the code pieces we > > have here and there (dl10n, amongst others), and the new ones we will > > develop to fully leverage the beast. > > Do you want to extend pootle with the l10n pieces and tools in Debian? > Contribute back to Pootle? Some feature has to be implemented within pootle, such as a reviewing mecanism (a way to not distribute a translation before it was seen and approved by at least N translators -- used to work in DDTP -- and maybe a bug tracking mecanism for translations) and mail interface (for the ones doing most of their translation in the Paris transportation system and thus off line and such). But on a larger plan, pootle would become one piece of the picture. Maybe something like having katie (or a friend) run dl10n on all uploaded packages to extract the stuff which need to be translated and push it into pootle. The extraction could be eased by po4a for everything beside regular program message catalogs. Then translations are feeded by translators into pootle, something could trigger an automatic upload to the archive so that the users benefit of them. These translations would use the override mecanism to get also added in ant further [manual] upload. Issues here are: 1) not to mess with maintainer work, ie don't erase translations provided by the maintainer with translations provided through pootle. Doing so only for packages containing a debian/dl10n file (and thus specifying that the maintainer wants so) could solve it. 2) technically doing an upload when translation is provided. You have to actually build a new package with the same content beside the translation, a new changelog entry and a version number, etc. Cryptographic issues may also arise. It's not impossible, but is still tedious. This tool could also help in l10n-only NMU rounds. > > We may eventually actually implement the dreams of a system automatically > > extraction all the translatable material from uploaded packages, and doing > > The itch is soooo bad that is impossible to think that nobody will not > scratch it. I'm scratching it since years, and I'm not the only one. Even if the big picture becomes clearer and clearer, with different software pieces maturating, I can't imagine that it will become reality before etch+1, actually. Bye, Mt.
Attachment:
signature.asc
Description: Digital signature