Modernizing the access to gcc sources ?

[Marc Espie <espie at schutzenberger dot liafa dot jussieu dot fr>]

I'm currently about to slurp a large amount of gcc.gnu.org bandwidth.
Namely, I have need of
- the gcc 3.0 branch,
- the gcc 2.95 branch,
- the gcc current branch.

Plus, gcc.gnu.org is very often a bit slow in answering diff/log requests,
even though that kind of stuff is very useful.

In the BSD world, where cvs is very widely used, we do have a few very
convenient services for these kind of queries.

Namely, all three BSDs allow one to slurp the whole cvs repository, through
at least sup mirrors, and very often cvsup mirrors.

When one is used to this kind of service, the way the gcc source tree is
handled looks positively antique. :-/
(see http://www.openbsd.org http://www.netbsd.org http://www.freebsd.org
for examples of such facilities in wide-spread use).

Setting up a sup mirror is rather trivial, it's just a question of saying
which files can be mirrored, and setting up an eventual access list.

I'm not saying gcc.gnu.org ought to provide such an access for everyone,
but I'm reasonably sure at least a few ftp mirror sites could provide
`first stage' sup mirrors (updated every 12/24 hours), that people could 
then use to grab a full read-only copy of the cvs repository.

I would even volunteer to set such a thing up, except for the fact that I
don't have any bandwidth I can use for that (my lab wouldn't vouch for it).

Note that checking out/updating a tree from several copies of the same 
repositories is about as simple as specifying cvs -d url   instead of plain
cvs.