Rosetta/bzr (was: What is an Upstream?)

[kiko at async.com.br]

On Wed, Nov 08, 2006 at 03:07:01AM +0100, Dominik Zablotny wrote:
> 1) Most of external data is stored in version control systems. It is
> easy to move data between them, but not between VCS and Rosetta - here
> we have one-way, central-navigated traffic. Extracting single files from
> Rosetta is *unnecessary* manual work.

Yeah. VCS integration is definitely a killer feature here, and there is
both good and bad news there. The good news is that it's planned and
having Bazaar means it's really possible to do it in a nice way, with
branches open just for translations. The bad news is that it's not here
yet, and needs to be fit into a future Rosetta development cycle. We're
getting there.

> 2) To correct single string noticed in application or reported by
> someone, you need in fact export and import template. Doing that in
> local .po files is many times faster and you can grep all of them at
> once.
> 
> 3) You sometimes need to do a bulk replace, with or without per-string
> review. Imagine how fast it is with Rosetta ;)

Yeah, that's crazy. Once VCS integration is there, though, it's a matter
of updating the file and committing to get everything in the web UI
again. We could also have a web frontend for mass-updates, if people
think it was justified (I'm a big fan of the VCS integration myself).

> 4) Translation teams often have their own dictionaries and string
> translation databases compatible with KBabel. There's no way to
> synchronize them with Rosetta, in any direction, or for example to use
> specific suggestions only for GNOME or KDE

Yeah, that's another item in the roadmap for Rosetta -- glossaries.

> 6) To do correct translation, it is often needed to run application with
> new file, which can be somewhat automated. Would new translators bother
> to import .po files, if they are not used to them?

That's a really good point. To test translations you just did is a bit
of work -- download, compile, install. There's no easy way around this
unless you consider daily language packs or something even more
advanced.

> Speaking less serious, my proposal to lower entry barier even more is to
> print templates on a dead tree, spread them through ShipIt together with
> Ubuntu CDs and collect translated sheets :-P

Who's going to do the typing?
-- 
Christian Robottom Reis | http://async.com.br/~kiko/ | [+55 16] 3361 2331