yet another dvcs shootout (this time python-devel)

[Paul Moore]

2008/11/6 David Cournapeau <david at ar.media.kyoto-u.ac.jp>:
> Can it be that network latency matters much more than bandwidth ? I
> noticed that when Georg Brandl compared hg and bzr, it got *only* 3-4
> times slower instead of several order of magnitude in my case. That's
> really too big a difference,

Quite possibly. What little I know of the protocols of hg and bzr is
that hg goes to significant effort to use streaming protocols wherever
possible (both over the network and on disk) - avoiding as much as
possible any "lots of little requests" behaviour. I don't know if bzr
makes any attempt to do this, but Matt Mackall (hg developer) claims
that this is where hg gets its big performance wins from.

Paul.