Debian i18n - Ideas for a common framework
In December 2005, motivated by the Extremadura WorkSession Debian i18n maillist started a discussion about some ideas and requirements to build a translation system. The idea is try to figure out the points where we can start working.
Requirements
- a way to authorize who is allowed to work on which translations
- a way for translation teams to decide how the review process of new translations is done ("needs 1 review", "needs X reviews", "does not need review")
- a way to detect consistency in translations (same term translated
- a way to automatically extract (from packages, CVS or SVN) and update the po's we use in Debian:
- po-debconf
- po-xml (some documents)
- po-ddtp?
- a way to work with things that are not po files:
- wml files
- some documents (XML or SGML)
- manpages
- a way for translators to disable Pootle-managed translations (I don't understand why, for example the Spanish translation of Gaim is in Pootle, I never asked for it and don't use it)
Considerations
- contributors should be able to work online and offline
- A lot of people talk about Rosetta and how easy is to contribute online. We should consider getting an interface ready for that, probably pootle will solve this. The concept of DDTP sounds interesting to the offline part, together with the pseudo-urls model, which brings the ability to avoid double work (ITT,ITR)
- package descriptions
Are we able to work in the packages translations (for apt-like tools) after a release? Or we should only work in sid? Could we integrate this "automagically" in the framework?
- unified statistics and reports
- Bring all the statistics to a central place, with reports of what is "on hold" (ITT,ITR), pending translations, outdated and manuals like D-I and DDP. We should really try to get everything together.
- helper tools
- Use of helper tools like dictd and spellcheck systems.
Ideas of how it could be
- Web based with PO download/upload
- Role based (administrator, translator etc)
- Process for a person to review individual strings and approve.
- Each string needs approvement from at least 1 person
- Status codes for each translation: alpha, beta, under review, approved.
Infrastructure
We are under discussion of how the infrastructure should work. Specially some of the interfaces between software and translators. One of the discussion points is about xliff format and how we could work with that. The core idea is base on the use of "po files". We need to found some agreement on that topic, you can check more on this thread.