Some unscientific timing results (on the Python source tree)

[David Cournapeau]

Paul Moore wrote:
> In my view, this is more "normal" behaviour anyway. It may be possible
> to speed things up, but in real world use, we have to live with what
> we get. Equally, I didn't do anything but use the repository formats
> that came by default (the Python repo tarball is rich-root-pack) and I
> didn't do any explicit bzr pack operations, etc.
>   

For this specific comparison with mercurial, note that mercural does use 
a "smart" server if you use http. If you want to compare the same thing 
for network operations, you should use http-static for hg url:

http://www.selenic.com/mercurial/wiki/index.cgi/StaticHTTP

Or use bzr:// for bzr if comparing against hg clone http://

I would say that network operation aside, the numbers you are giving 
seem reasonable (that is, in my experience, for most operations, bzr 
st/diff/commit/push/pull is around twice slower locally compared to 
mercurial for medium repositories with a few thousand revisions; log is 
much slower, but that's something which is being worked on I think ?).

cheers,

David