Eficiency [Re: comparison with mercurial?]

[Martin Pool]

On 28/10/05, Jérôme Marant <jmarant at free.fr> wrote:

> >From the same source (the IceWM source), with mercurial 0.7 and bzr 0.1.1
>
> - The initial commit of sources (after adding all files) is 4 times faster
> with mercurial
> - Local branching of sources is 6 times faster with mercurial
> - Branching via HTTP (old-http from mercurial) is 3 times faster with mercurial
>
> That said, I'm still interested in bzr.
>
> Are there some hopes to see bzr efficiency improved anytime soon, and how?

Yes, there are.  There are three main efforts in this direction.

The first is in switching to a better weave-like format, called a
'knit', which has similar append-only update properties to hg.

The second is to store all revision files within a single container,
which should drastically reduce the number of http roundtrips we have
to make, for the bzr.dev branch from about 2700 to about 300.

The third is to do overlapped http requests, so as to avoid the
latency that comes from naive http against a passive http server.

There are some other things but these are to my mind the main ones

--
Martin