[gentoo-dev] R/O CVS access and its purpose for ATs (was Email subdomain)

[Daniel Ostrow]

On Tue, 2005-11-22 at 17:56 -0600, Lance Albertson wrote:
> Marius Mauch wrote:
> > On Sun, 20 Nov 2005 09:32:55 +1100
> > Ben Skeggs <darktama@×××××××××.au> wrote:
> >
> >
> >>Anyway, the most important reason for the GLEP (IMO) is giving AT's
> >>r/o access to CVS.  When working on bugs, it's always fun to find out
> >>that the problem has already been resolved and just hasn't made it to
> >>your local rsync mirror yet..
> >
> >
> > Out of curiosity, what's the more important aspect of r/o cvs:
> > - more up to date
>
> Not necessarily true. We would not have the anon cvs access from our
> primary cvs server. It would be synced on a regular basis to a separate
> box. The newer cvs (which isn't on lark yet) may give us capabilities to
> have a more 'live' cvs anon system. But as of now, the best infra can
> provide is 30 minute updates. I don't want to poll the cvs more than
> that to keep down the load.
>
> > - easier selective updates
>
> Yup, that's definitely a plus.
>

And herein I think lies some confusion. Personally if I were an AT both
would be important but more to the point the "more up to date" issue
would be the most important. I think that there is a need for the ATs to
be able to work in direct conjunction with a dev, an AT catches an
error, a dev fixes it in CVS using a *well tested* patch, an AT does a
`cvs up` and retests to try and catch *other* errors all within a matter
of *single digit* minutes. This is a very powerful tool, rather then
what they have to do now which is either wait for it to hit the rsync
mirrors, a dedicated rsync mirror, a dedicated anoncvs box, or e-mail
the ebuilds (and patches) back and forth. Note the two highly stressed
things up there...this should not be used so ATs can vet patches (wither
to ebuilds or to source), the patches should be well tested long before
they reach our tree...

Lance:

I know this is a far cry from what you are proposing, and I understand
that the present CVS server cannot handle this sort of load but I
believe that this was the original intention at least...someone correct
me if I am wrong.

I think that this issue has to be nailed down *before* we get any
further in discussion.

-- 
Daniel Ostrow
Gentoo Foundation Board of Trustees
Gentoo/{PPC,PPC64,DevRel}
dostrow@g.o

-- 
gentoo-dev@g.o mailing list