Bazaar-NG vs. Mercurial -- speed comparison

Bryan O'Sullivan mambo.chicken at
Thu May 18 20:36:35 BST 2006

On 5/18/06, John Arbash Meinel <john at> wrote:

> > It actually works over plain http, though you can't push over HTTP yet.
> I'm guessing that it would be slower over plain HTTP, though I obviously
> don't have any actual tests to prove that.

If you don't have a CGI server, but are pulling from plain files, it
is slower, yes. With a CGI script, it's pretty quick.

> That said, representing things in memory as weaves does impose some
> overhead. But knits do provide support for only reading a little bit,
> rather than having to read the whole thing.


> Well, your clone benchmarks are all for hg using hg, and bzr using bzr.

Yes, my brain fell out into my lap when I wrote that. Sigh.

> But I will also say bzr is slow to create working trees. For mercurial
> are you actually extracting the text from the repository, or are you
> just copying the files on disk?

Yes, we extract the files from the repo. Copying the working dir is
messy, because you have to be sure to either not copy or reconstruct
all of the stuff that the user has modified.

More information about the bazaar mailing list