From: Walter Landry (wlandry_at_[hidden])
Date: 2004-01-26 20:13:33
Vladimir Prus <ghost_at_[hidden]> wrote:
> Walter Landry wrote:
> > Perhaps this isn't the best time to raise this, but wouldn't it be
> > better if Boost eventually moved to a more distributed version control
> > system?
> I don't think that's going to help. Distributed version control system allow
> you to commit to your local repository and then push the changes to the
> master copy. That's probably nice for some cases -- the favourite example
> of proponents is hacking in a train/plane. But for Boost, in order to test
> the release, you *still* need to push all the changes to the master copy --
> which still remains single point of failure.
> I'm not sure if existing system will allow to collect the changes on
> machines used for regression testings -- but even if so, that would be
> really inconvenient. Imagine that if you fix a bug, you'd need to push it
> to all machines used for regression testings. Or otherwise, all testers need
> to pull changes from all known local repositories.
One of the problems people are currently experiencing is that you
can't do any commits when the server is down. A distributed system
will allow that. But as you noted, you also have to be able to
publish your results. With a distributed system, you can publish and
coordinate from anywhere. For example, Rene Rivera might tell
everyone on this list (and the regression machines) to look at Rene's
own repository (or a mirror that someone with more bandwidth put up)
instead of the official Boost.Jam repository until the main server has
But like I said, I don't actually know of any systems that will cover
_all_ of Boost's requirements. At least as I understand them.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk