Re: Broken po files

[Malcolm Tredinnick <malcolm commsecure com au>]

On Wed, 2003-09-10 at 11:24, Carlos Perelló Marín wrote:
> El mié, 10-09-2003 a las 02:43, Malcolm Tredinnick escribió:
> [...]
> 
> > As far as Scrollkeeper goes, it is intentional that we are not forcing
> > everybody to use UTF-8 when creating their translations.
> > 
> > Somebody is going to have to explain to me where the breakage is here,
> > since I do not see what is going wrong: translated strings are not piped
> > directly to applications, they go via an XML file with valid encodings.
> > Internally, we are not calling bind_textdomain_codeset(..., "UTF-8") or
> > anything like that, so things work consistently under the covers as
> > well.
> 
> If the .po strings are only for .xml files, there is not problem (as far
> as I know about it)
> 
> > 
> > Scrollkeeper has to work in areas where the native encoding is not
> > UTF-8, so we have striven to keep it portable in this fashion.
> > 
> > Are you sure this is not just a mistaken over-generalisation from the
> > fact that GNOME applications all use UTF-8 under the covers to ease
> > things like GTK+ integration? Scrollkeeper does not have that
> > requirement.
> 
> Glib and Gtk >= 2.0 needs only UTF-8 strings and the problem is when you
> have an old libc version (I don't remember the release when this
> changed). With old glibc, if you use nonUTF-8 .po files, all
> application's strings will be nonUTF-8 a then, gtk/glib will give you
> lots of warnings/errors and you will have problems with your
> application. With the a recent glibc, the library recode the .po file
> directly to your own locale encoding so you need the 
> bind_textdomain_codeset(..., "UTF-8") call to prevent that it use
> nonUTF-8 encoding if you are using, for example, es_ES.ISO-8859-15 as
> your locale.

Yes, thanks. I know all this. In the case under consideration, since
scrollkeeper does not use GLib (mostly for historical reasons), there is
no problem here (that is what I was hinting at in the first email).

I suspect the original IRC comment was just based on somebody grepping
through files and not appreciating all the subtleties, rather than
having seen actual problems. (Don't get me wrong: it was a useful
original mail, since most of the packages were problematic)

Cheers,
Malcolm