Re: Free Translation Project

[Tom Tromey <tromey at cygnus dot com>]

>>>>> "Martin" == Martin v Loewis <martin@loewis.home.cs.tu-berlin.de> writes:

>> The obvioius problem is what do you do about front ends that aren't
>> released by us?

Martin> It's not clear whether these front-ends will support a message
Martin> catalog at all; if they do, we should advise them to use a
Martin> different textual domain (e.g. "gpc" instead of "gcc").

This means either we must use dgettext in the gcc core or (e.g.) the
gpc maintainers must copy the gcc po files.

My understanding is that gcc is slowly separating the front and back
ends.  I'd like to see this reflected in the translation domains as
well: the core compiler would use dgettext and have their own domain,
and each front end could have its own domain (used by default).

To me this seems cleanest.  It also avoid problems with people
updating the po files but not having the Chill compiler checked out.

One idea would be to put a po directory into each front-end's
subdirectory.  This is similar to the approach I took in binutils.

Tom