[darcs-users] peer to peer darcs

Andrew Pimlott andrew at pimlott.net
Fri Jul 30 20:26:07 UTC 2004


On Fri, Jul 30, 2004 at 03:12:09PM -0400, S. Alexander Jacobson wrote:
> On Fri, 30 Jul 2004, David Roundy wrote:
> > The order in which patches are received wouldn't be a problem unless they
> > get a patch bundle before they have its context,

Isn't the obvious partial solution to this to send out patches with a
minimal context, instead of the full context as now?  Or maybe it should
send both, and give apply a --force option to use the minimal context.
My understanding is that the context in a send is the intersection
between the two repositories.  This is conservative, but generally much
more than is needed.

> This architecture struck me as viable because I
> interpreted the "theory of patches" to mean that
> darcs could cope with patches arriving in
> different orders:

It can.  What David is saying is that if the send was generated against
a newer repository than you have, darcs will not let you apply it.  But
I think this is just a matter of caution, rather than a real limitation.

> > The easy (and probably best) way to do this would be to have each person
> > have two repositories, a mirror of a virtual "central" repository, and
> > their own working repository.  They would never make changes by hand to the
> > "central" repository, but instead would send changes to the list and have
> > those changes automatically applied to their "central" repository.

If you could send minimal contexts, you wouldn't need a mirror of the
central repository, would you?  (Except to know which patches to push,
but that wouldn't be hard to keep track of manually.)

Andrew




More information about the darcs-users mailing list