[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Fixing d-c l10n bugs and SVN/CVS commit access

On Tue, Dec 05, 2006 at 06:48:48AM +0100, Christian Perrier wrote:
> > Fixed dictionaries-common package (0.70.10) uploaded to sid this morning
> > with medium urgency.
> Agustin, does this version include all the pending debconf l10n
> fixes ?
> dictionaries-common appears on the l10n NMU page which lists all
> packages with longstanding l10n bugs. I haven't lookes closely to it,
> though.

AFAIK, there is only one, [#369172 cannot support internationalized debconf
names], and this is not an outdated translation, but something by far more

For debconf question about default ispell dict/wordlist, values are gathered
from info provided by (to be) installed ispell dicts/wordlist packages,
where the unlocalized debconf entry for that dict is also used internally as
an unique identifier. If those entries are localized, the returned value does
not match the identifier, and things get broken. For that reason we use some
sort of poor man localization, where names are as

local name (english name)

where local name is in the dict language, and entries are no further
localized. That should serve the vast majority of sysadmins, because the
name is both in english and in the local language (or in the usual
transliteration, at maintainer will). What the bug report asks is to also
make that entry localizable, and that is what is not at all trivial. I made
some experiments with the original package/languages entry for each dict
package and a companion package/elanguages entry which could be localized,
so package/languages is used as unique identifier, but question is modelled
after package/elanguages. However this needs to

For every (to be) installed ispell dict /wordlist,

a) Get package{e,}languages for each dict package
b) Check if question needs to be asked.
c) Map values from one to other and from other to one
d) Get current shared question value and obtain its elanguages mapping.
e) Feed shared equestion with possible elanguages values
f) Feed equestion value with the elanguages equivalent for the shared
   question value.
g) Ask equestion.
h) Feed question value with the languages equivalent for the shared
   equestion value.

No need to say that I dislike playing with debconf this much. Also, this
translations should go to every ispell dict/wordlist, putting work on
translators for something I think is of low priority for them, because
current setup serve most needs. Since this can open a can full of bugs,
will not consider it at all for etch, and I tend to think that this change
does not worth the effort even later.
> Actually, as d-c is listed in D-I "level5" packages, which means you
> get a lot of debcofn translations for it, I think it would help if I
> can get SVN commit access to your repository to be able to commit the
> incoming translations and save you time...

Thanks for the offer. Fortunately the updates rate was already high in
sarge, so the current rate is low, something we can easily deal with.


Reply to: