[Po4a-devel][CVS] po4a/lib/Locale/Po4a Xml.pm,NONE,1.1

Denis Barbier barbier@linuxfr.org
Tue, 20 Jul 2004 10:17:07 +0200


On Tue, Jul 20, 2004 at 12:34:51AM +0200, Jordi Vilalta wrote:
> On Mon, 19 Jul 2004, Denis Barbier wrote:
[...]
> > This would require a comment being added to PO files to let translators
> > know how to deal with this msgid, and my experience is that translators
> > hardly read such comments, because they use tools which do not display
> > them (kbabel, gtranslator, poedit, etc, very few of them handle comments
> > in a useful way).
> 
> I've been using poedit and it shows them well.

There are 2 kinds of comments: those extracted from sources, and
translator's comments.  IIRC poedit was messing up those comments,
there were discussions on Debian lists few months ago about this
bug.

> But I agree with you on the 
> translators behavior about the comments.
> 
> > This attribute may be automagically added, e.g. including translations
> > from xx.po could set lang="xx".
> 
> With this you're forcing the name of the po file, not much interesting in 
> some cases. And there's also another issue: there can be a language 
> without the translated DocBook templates, and they may want to set the 
> lang attribute to any arbitrary language.

I do not see your point here, sorry.  Note also that AFAICT your
solution below does not handle that case.

> There's also the possibility to have more than one lang attribute
> (with different values) around the document.

They can be created by reading all *.po files.

> I think that another solution would be to create artificial msgids:
> "TRANSLATE ONLY AFTER THE ':'\n
> THIS IS THE VALUE OF THE aaa ATTRIBUTE OF THE ttt TAG: en"

I tried this approach, and some translators did translate those msgids ;)

> Maybe this isn't very friendly to pure translators, but I think that 
> software translators should have a minimal idea about the format they're 
> translating (see the 'E<lt>' stuff in the extracted strings from the man 
> pages).

I slightly disagree, my experience with po-debconf (a Debian specific
tool) is that such tricks confuse some translators, so the question is:
do we want a l10n framework usable by everybody, or by experienced
users?  IMO extra complexity should be added only when there is no other
choice.

I did not yet read the thread about the 'E<lt>' stuff, but these escapes
appear in msgids, this is different from diverting msgids from their
original role (ie. having msgid/msgstr handle original and translated
strings).

While we are talking about complexity, plural forms is a very hard
topic, maybe it should be discussed too.

Denis