Public SourceMage Discussion List

[Hamish Greig <hgreig AT bigpond.net.au>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Tue, 16 Sep 2003 01:41, Eric Sandall wrote:
> I like this idea better than posting a devel tarball, but I also wonder (as
> has been mentioned by Dufflebunk) about the load this may put on the P4/CVS
> server? Would it be too much bandwidth?  Too much CPU load?  Perhaps the
> easiest way would be to just post the devel tarball to codex.sourcemage.org
> so that we can make some people happy (this is, after all, about choice, so
> if people want the devel tarball, we should provide it) and not overload
> our development servers.

Removing the devel tarball has had two great benefits, bugs are now fixed in 
test and test /stable grimoires are more uptodate than ever before.
I am extremely worried that reproducing the devel tarball and publishing it 
will have a negative affect. With devel tarball available again how long will 
it be before we have bugs only fixed in devel while the guru awaits some 
feedback before integrating? Will it be a week before integration ? 2 weeks ?
As far as i am concerned the only differences between devel and test should 
be 
major version changes of integral spells like gcc, glibc etc (ones that the 
whole system relies on, or that many spells depend upon) and "Works in 
Progress" spells that are getting rewritten or added that aren't working yet 
and will /could break the users system.
In my opinion, all bugfixes should go straight into test. NO questions. If 
for 
some reason it shouldn't go into test then it is arguable whether it is a 
valid bugfix.
Devel should only differ from test in the ways previously mentioned, and to 
make matters even simpler i think it shouldn't even be a complete grimoire.
I think if a spell in devel is the same as in test in should be removed from 
devel repository . If a spell is unique or different it stays. If people want 
to do major rewrites like the grub spell it is an easy thing to branch the 
test version into devel (with revision history thanks to perforce) and then 
it can be worked on. CVS "branching" loses the revision history as i 
understand it so it isn't an issue here.
This would also reduce current guru's workload slightly by not having to 
"integrate" every fix. They would be fixing only in test.
The perforce server "might" be able to handle "anonymous" access for a 
grimoire with only 1-2 dozen spells in it and cvs would just be a very small 
devel module.
I think publishing the devel grimoire as a complete tarball will start a 
clock 
ticking, and it will only be a matter of time before we are in the same 
position as two months ago, a bug riddled test grimoire and a months out of 
date stable grimoire.
I think this is something that the people advocating devel tarball 
repackaging 
are failing to understand. If i could think of another way of explaining it i 
would, but i have tried until i am blue in the face.
Hamish
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)

iD8DBQE/ZpPU8fSufZR6424RAmJ0AJ9MccCQlg9FAzJpZP16W/zJhFprwQCeL9TN
Awd2Q2dxyAml4td+nRnIadY=
=Vpwc
-----END PGP SIGNATURE-----