Shallow partial clones: how to proceed?
Matt Mackall
mpm at selenic.com
Sat Jul 19 10:49:29 CDT 2008
On Sat, 2008-07-19 at 09:28 +0200, Peter Arrenbrecht wrote:
> So I've had a chat with mpm and cmason on IRC about this. I going to
> pursue my approach (I'll call it parent nulling) as mpm thinks both
> parent nulling and cmason's punching have their merits. Key points:
>
> * We need to rethink the wire protocol and bundle format so we can
> pass deltas based on something else than prev, and rev flags. The
> delta thing would be not only for parent nulling, but also for being
> able to send deltas against a real parent when prev != parent0,1.
>
> * mpm would like the new wire protocol to get rid of the duplication
> with localrepo.changegroup() and localrepo.changegroupsubset(). Tough
> one.
It's not inherently tough as one is a proper subset of the other. But
changegroup is faster as it does less work and that's a nice property to
keep.
--
Mathematics is the supreme nostalgia of our time.
More information about the Mercurial-devel
mailing list