Re: Road map--or some such.

[Wolfgang Jaehrling]

On Mon, Apr 19, 2004 at 07:04:14PM +0200, Alfred M. Szmidt wrote:
> The package management system should of course support binary packages

Yes, yes, yes!

> I was intending to use GAR (by Nick Moffitt), it is quite small and
> [...] from the looks it is easy to add features too.

Sounds like the kind of thing we want.

> I am not particular fond of dpkg/rpm, since they provide to much, and
> are to bulky for what they do.

Yes, they are pessimal (opposite of "optimal" :->).  Complicated
software is a Bad Thing.  On the other hand, using them will be less
work than writing a new one from scratch.  Maybe it is feasable to use
an old version of dpkg which is not yet that bulky?  Actually, I like
this idea.

> Oh, I might note that I want to see a working system (doesn't need to
> be in a releasable state) by the end of this year.

A few more things:

1.) Did you think about what software we want to include?  I strongly
favor to not include thousands of redundant window managers, editors
and e-mail clients (and whatever).  It is enough to include the most
important packages; of course we have to prefer GNU software, so we'd
use GNU Emacs rather than XEmacs, gcron rather than Vixie cron.  In
some cases, we might also want to provide additional non-GNU packages
if they are very popular.  Does this sound like a reasonable policy?

2.) Another question is whether to keep the base-system small and if
we find that important, how to do so.  For example, including Perl
(which is necessary at least for GNU stow i think) and Guile (which
will be necessary if we use gcron, and if we end up using dmd)
somewhat "bloats" the base system.  Personally, I don't think it's
essential to have a very small base system, because a) hard drives are
large these days, so it does not really matter b) tools exist so that
we can use them; avoiding them even if they are useful for us does not
sound overly intelligent.

3.) What do you think about having all relevant projects officially in
the same package (like it is done for coreutils)?  Then it would be
less work to assign copyright to the FSF (stupid reason, I know, but
still...).

4.) I propose the codename "Liberation" for the first release of GNU.

Cheers,
Wolfgang

-- 
Repeating false statements makes them true.
Repeating false statements makes them true.
Repeating false statements makes them true.