Re: other/2857: i18n, translations does not work

[Dennis Bjorklund <db at zigo dot dhs dot org>]

On Thu, 24 May 2001, Zack Weinberg wrote:

> On the other hand, it might be more useful to have a known place to
> work from, which would argue for leaving gcc.pot in CVS and updating
> it periodically.

But you dont ever want to have a known place to start from that is not up
to date. Then you would just translate strings that would never be used
anyway since they are removed or change in the source (but still in the
gcc.pot since it's old). You always want to update your .po file when you
start translate.

And you can not have some automatic update of the .po file, since they are
checked out by the translator and if it's updated while he is translating
then there will be a conflict when he tries to commit his changes. So you
don't want to update during translation.

All other projects I've seen keep the .pot file out of the cvs and lets
the translators update their .po files themself.

I also think the fsf way of having a central place where you put .pot
files to translate is broken for big programs. Translations needs to be
tested and there is usually a lot of errors in the code that one finds
first when one do translations. The last week Zach found 1000 more strings
that where not marked for translation before. And there are a lot of
constructs in gcc where the output string is created dynamically in gcc
that only works in english. You don't see the problem until you start to
work on it. And to wait until a release and to put a .pot file on the fsf
ftp site and hope it will work is too much to hope for (at least for big
programs like gcc). Also remember that a gcc translation is a 400k file
with almost 4000 strings to translate.

I don't see why the .po files should be treated any different then other
source files. They need to be updated incrementally, and for that cvs is
usefull.

-- 
/Dennis