| Thread Previous • Date Previous • Date Next • Thread Next |
On 2010-10-23 12:21, David Planella wrote:
I think what we could do as well, though, is when we drop a release from exports, we do a final, one-off, copy to -proposed, call for testing and upload to -updates. This way, even if the release is no longer actively translated users can still get the benefit of message sharing, obtaining new translations from the newer releases.
All this makes me wonder if delta language packs still make sense. I don't think the code that produces them "knows" about message sharing, so there may be tons of changes that don't make it out to the users until the next full language pack.
If that is the case, maybe we could make our users a lot happier simply by ensuring they really get the translation updates. We could probably fix the deltas, but:
* They'll grow faster and so will save us less processing time. * In Launchpad at least, full exports are no longer very costly. * We've realized we don't want quite that many exports anyway. * For syncing our 2 cron jobs, "consistent" is better than "fast."So does the complexity of supporting delta language packs still carry its weight? Purely on the Launchpad side I would say no. What does the picture look like once the tarball leaves Launchpad?
It'd be nice to get translations to the user much more quickly than we do now, but do deltas speed that process up very much? Certainly not the part before the language-pack-o-matic stage, because that job runs at a fixed time. There may be other ways to reduce download size and shorten the path to the user. Arne was thinking of letting users get them straight from Launchpad.
Jeroen
| Thread Previous • Date Previous • Date Next • Thread Next |
