Re: ITP: gettext-base
On Tue, 29 Jun 1999, Julian Gilbey wrote:
> > > No objections from me. Sounds fine. As a harder project (of whose
> > > worth I am uncertain): figure out how much more space could be saved
> > > by splitting the catalogs as well.
> >
> > If the current scheme is well enough for you, please don't think of
> > splitting the catalogs. It will create additional problems on the
> > translators's side.
>
> No, it won't. I wouldn't dream of having more than one translation
> file, but there will be a Perl script or similar which will automate
> the splitting of the catalog file. My idea is to create
> gettext-base.{po,mo} files which contain only those messages needed by
> /usr/bin/gettext itself. The main gettext.{po,mo} files will continue
> to contain everything. Should be very easy to implement. The only
> thing which needs thinking about carefully is how to make sure that
> gettext finds the catalog. gettextp.c could be modified to read:
>
> /* Set the text message domain. */
> bindtextdomain (PACKAGE_BASE, LOCALEDIR);
> textdomain (PACKAGE_BASE);
>
> and then a configure-time option could select whether to have
> PACKAGE_BASE equal to ${PACKAGE} or equal to "${PACKAGE}-base". In
> the latter case, both catalogs would be needed in order to make
> gettext and xgettext (say) both work.
> [...]
Well, now that I've already uploaded gettext-base and it has only 70K of
size (which I think is quite reasonable), I must say that applying an ugly
patch to gettext source is not very appealing and I'm not willing to do it
just for fun.
I will repeat: There are more packages in the base system which are much
larger than this. If the upstream author does not like the idea (which I
agree it would be nice) of including the gettext program into the libc
sources, I would suggest to leave things as they are now.
Thanks.
--
"129d69bb7b1bff181b54ae0a03ebc730" (a truly random sig)
Reply to: