Re: GNU gettext-0.10.36 for non-ASCII msgid

[Keld Jørn Simonsen <keld dkuug dk>]

On Mon, Apr 09, 2001 at 01:00:03PM +0200, Pablo Saratxaga wrote:
> 
> The problem is: in what encoding would be the strings in sources?
> iso-8859-1? koi8-r? jis0x208? utf-8? depending on the main author?
> depending on the release date?
> 
> IMHO the sources must have only ascii.

Yes, that is the most practical thing for the moment.
I recommended to use \uxxxx notation in a previous mail.

> The problem is that until *everybody* switches to utf-8 (authors and
> translators) the msgid entry in the po files will cause problems, as it may
> be changed by the program used by the translator and so the corresponding
> msgstr may point to a differently encoded msgid; that is, the string won't
> be translated. I have already the case and I have to manually correct
> such problems with some xmms translations as the sources include 8bit
> iso-8859-1 chars.

Another thing is that current gettext allows for fallback, so if
you have a character that cannot be represented in the encoding
of your execution locale, then an approximation of it will
be displayed. I have not tried this, eg with a \uxxxx entry
in a msgid or msgstr, but it should all be in the tools as
currently distributed.

Keld