[Caminol10n] Localization Application

[Smokey Ardisson]

At 12:33 AM +0200 on 6/28/07, Marcello Testi wrote:

>Il giorno 27/giu/07, alle 22:42, Martin 'MMx' Creutziger ha scritto:
>
>>Before you reinvent the wheel here, have you taken a look at the 
>>"L20n" concept
>>Axel Hecht AKA Pike presented as a possible future l10n solution for Mozilla
>>products? Maybe it is usable for Camino, too.
>
>I've just looked around for the first time at the l20n thing and I 
>think it addresses quite different needs than those explained by 
>Tobias: while L20n aims at bringing more efficiency making the 
>translations less linguistically "dumb", the trouble we face here 
>has two sides, both with little to do with grammar and vocabulary.
>1) a tool that must be reliable in identifying and extracting all 
>untranslated strings in the application so they can be translated 
>without wandering through interface bits

Perhaps JJ has an update on the status of Apple releasing a 
non-broken version of AppleGlot?

>2) a way to distribute the work among l10n teams while keeping all 
>together chronologically

What did come to mind when MMx mentioned Axel's work is the 
compare-locales script 
<http://mxr.mozilla.org/mozilla/source/testing/tests/l10n/scripts/compare-locales> 
It's not usable at all for us (since it works on Mozilla jars and 
their dtd and properties files), but long-term it would be useful to 
have an equivalent tool for Camino.

My thoughts are it would do something like this:

1) Invoke whatever command-line magic AppleGlot uses on the files in 
Resources/localized/*.lproj and PreferencePanes/*/Resources/*.lproj 
to produce all the individual temporary .strings files
2) Compare these (plus the actual .strings files, and the ReadMe and 
Release Notes) from the English.lproj with a given localization
3) Report on untranslated strings (strings whose text is equivalent 
to the English string) and "missing" strings (strings present in the 
English version but missing from a localization, meaning that a nib 
or .strings file is out-of-date in a localization) and untranslated 
lines (paragraphs) in the ReadMe and RTF.
4) Repeat with every other localization
5) Publish results to the web

The prerequisites (besides a good hacker to write the script) are 
landing the localizations in *some sort* of VCS (not necessarily 
cvs.mozilla.org) and then a Mac to run the tool on a regular basis, 
pulling English files either from cvs.m.o or from the latest nightly, 
and localizations from the localizations VCS, running the tests, and 
then posting some results to the web.

Anyway, just an idea for discussion.  There are several hurdles 
there, but if people like the direction (or come up with better 
ideas), it might be something for some of the caminol10n community to 
work on while development of Camino 1.6 is underway and not yet ready 
for substantial localization work.

(Another idea, to solve a different problem, is to simply create a 
list of all the files that get localized, and have a script poll 
cvs.m.o or bonsai.m.o periodically to see if any have changed, which 
could do the simple task of alerting everyone when we change a 
localizable file without any human communication required.)

Smokey
-- 
Smokey Ardisson
Co-Lead
Triage/QA and Website & Documentation
The Camino Project