On Mon, May 02, 2005 at 07:26:39PM -0400, Joey Hess wrote: > Christian Perrier wrote: > > Well, my analysis was that measuring this is far more difficult with > > the netboot image because as soon as the choose-mirror step is over, > > the installer downloads udebs from the archive....and of course, those > > ones are not including the new language. > > > > Or do I understand wrongly? I'm not sure I understand properly what > > you mean by "the very end of anna run", sorry for that...I'm still > > learning things about D-I..:) > > Right, I mean the very end of the download (and install) process for > udebs is where you need to measure to get the full possibly memory > impact of a new language. Of course that's much harder than just adding > a language to the initrd. Isn't it possible to build a translation-only udeb, with all the translations for all the udebs? It would be temporally (if not, the installer becomes kind of monolitic again), but it may help in this kind of computations. And, concerning the l10n space consumption, what about pruning extra translations when udebs are extracted? When flaming me for proposing already discussed ideas or other reasons, consider that I'm not on -boot and may have missed something ;) Bye, Mt.
Description: Digital signature