Not a holy war - just some salient facts

Tom Widmer tom.widmer at
Mon Apr 12 07:07:05 CDT 2010

Masklinn wrote:
> On 9 Apr 2010, at 10:28 , Adrian Buehlmann wrote:
>>>> But it seems a lot of users never check their repos. Otherwise I can't
>>>> explain why anyone ever would use bzr.
>>> So, are you saying it has other problems, making it a poor choice?
>> I can't tell if it has other problems or not. I stopped looking at it
>> because of the poor bzr check performance, because reasonable verify
>> performance is a must have for me before even starting to consider using
>> a DVCS.
> In my experience (been using bazaar since November because employer uses it,
> previous experience was mostly with mercurial and darcs with a bit of git)
> the slowness you experienced with bzr check exists throughout the software
> (on some operations, including local-only, bazaar manages to feel slower
> than subversion).
> bazaar just feels slow to and through, and I'm talking bazaar 2.0 & 2.1
> using the latest and greatest repository format (2a). It was apparently
> even worse before.

> I'm sure there are others, but I have to go back to work. Anyway all in
> all after ~5 months working with it I'm *not* fond of bazaar compared
> to the other DVCS on the "market".
> Your mileage may vary, I hear others have less issues with it, I don't
> know what experience they have with git, hg or darcs.

Bzr has a few features that hg lacks:
- shallow clones (aka stacked branches)
- efficient renames/copies (and full rename tracking)
- switchable branches (hg can do this for named branches, but not easily 
for clones)
- Subversion branches (i.e. you can treat a Subversion branch much the 
same as a remote bzr one)

The architecture seems much more heavyweight in Bzr, which is perhaps 
not a good thing (and is probably the reason for the performance issues, 
though I'm sure it must have benefits as well). I believe the bzr 
codebase is at least 2x as big as HG...


More information about the Mercurial mailing list