Re: Administrativia

[Behdad Esfahbod <behdad cs toronto edu>]

On Fri, 19 Aug 2005, Bruno Haible wrote:

> As a result of last week's discussions, it's becoming clear to me that
>
>   - gettext's glocale is a good starting point for this project. The
>     LGPL license is apparently OK. There's a large overlap of functionality.
>     And the architecture (runtime + a cooker) fits as well.

Yes, we are going forward with LGPL, which makes it easier to
develop the library, since we can borrow code from almost all
interesting projects.


>   - The project - with the support for CLDR and the setter APIs - is
>     larger than what I had initially thought. As a result, I'm reconsidering
>     the infrastructure.

That's what I've been thinking about too.  I've not still got the
time to reply to that thread though.  But I'm trying to come up
with a map of completely isolated components, that may even have
different licenses.  For example, the whole user-configurability
should be on top of the basic functionality, and on top of that,
gconf should be a detail.  On the backend side, It may be
possible for example to have the CLDR wrapper with a wider
license, most probably borrowed from ICU.  I'm also thinking
about Unicode Character Database and regular expression
engines...


> * Who volunteers to contribute code or documentation to the project?

I'm in business from the October or November time-frame, but I
have hard deadlines before that.

> * Can it be a GNU project or part of a GNU gettext?
>
>   I would very much like it to be a GNU project, so that GNU gettext can
>   use it. (msggrep for example is hardly usable, because we don't have
>   a gl_regex function yet. glocale will fix this.)

That's what I was going to suggest.

>   When it is part of GNU gettext, all contributors need to sign papers
>   giving the FSF the copyright over their contribution. If it's a separate
>   project, the contributors can keep their copyrights, but it's harder for
>   the project to play a central role in GNU.
>
> * Should it be part of GNU gettext or separate?

I have another view of it, what about GNU gettext becoming one
part of the project?  I mean, gettext is probably quite smaller
in the scope and complexity, compared to the whole picture we are
building here, right?

>   Should it have a distribution directory, a web site etc. of its own?

I think it doesn't make much difference in the first days,
whenever we are near having something good to offer, we can
decide where to settle.  For now we need a CVS repository and a
bugzilla maybe.  Both are available readily at GNOME.  I cannot
figure out where gettext is developed.  Apparently not Savannah.
About becoming a GNU package, again, that would be easy enough to
do at any time, you know better.

>   The link between glocale and gettext is quite weak: just the 4 functions
>   from glocale/libintl.h. But:
>
>   If it's part of GNU gettext, it will find its way into Linux distributions
>   rather easily. Otherwise, it will be GNOME which "drags" it into the distros.

I don't think so.  Have you seen PCRE?  PCRE is a project we
should model after.  It started as a pet project for somebody
developing a simple standalone Perl-compatible regexp engine.
Now that's being used in Python and PHP, and GNU grep has an
option to use that.  Of course, it's crap compared to GNU regexp
engine or Perl's, but it's portable and standalone, and that's
what matters for many widely used projects.  What I had in mind
was pushing ourselves into PCRE, since it lacks localization
currently, so for example word boundaries do not work beyond
ASCII...  The only thing we should be extra careful about is
portability and size, and they would require us like hot candy.
:)

>   If it's part of GNU gettext, the sharing of some libintl files and
>   other infrastructure is easier.
>
>   It it's part of GNU gettext, it will be easier to install on non-glibc
>   systems. If it's separate, people will need to build and install
>   1. gettext (for the tools), 2. glocale (for libglocale), 3. gettext again
>   (so that msggrep works better).
>
>   What's your opinion?

I think we need a gettext inside glocale for the least.  One that
is locale-aware and all.  What this means, I don't know.  It may
be a syntactical inclusion of GNU gettext source in it, or a
the projects merged.  My primary concern is that the sudden jump
in the size and complexity of GNU gettext may not be as welcome
as you may think.

> * What shall be the name of the library? Is "glocale" OK with everyone?

glocale is of course the mechanical name.  I prefer something
smarter :).  We better not use the gl_ prefix BTW, suggests
OpenGL way too much.  "GLU Locale and Unicode support library" is
another geeky choice ;).  Everybody look for a good name.


> Bruno

--behdad
http://behdad.org/