Bazaar still below the radar when evaluating VCS tools

[Paul Moore]

On 26 February 2010 17:57, Parth Malwankar <parth.malwankar at gmail.com> wrote:
> One thing I think that has hit bzr is the "its slow" tag. While this
> is no longer the case this tag has stayed. Unfortunately there are
> no recent speed benchmarks available. Most are quite old. Probably
> users of bzr who have a blog could consider posting some benchmarks.
>
> I see some space benchmarks at http://wiki.bazaar.canonical.com/Benchmarks
> but there is nothing on performance. Maybe some performance benchmarks
> can be put based on some well known medium to large size projects already
> being handled by bzr. These need not necessarily be "bzr vs X" and could
> be just bzr numbers. It would give people evaluating DVCSes a feel for bzr
> performance.

The trouble is that the impression isn't based on comparisons and
benchmarks. It's a gut feel type of thing.

Just to see, I tried my (hugely simplistic, completely unfair, ...)
initial clone test - just create an initial branch of the Python 3k
repository using Mercurial and Bazaar. It's intended as a view of a
newcomer's initial impression of the 2 systems - no clever workflow
setup, no performance tweaking, just install the tool, search the web
for the place to get the Python source in the appropriate format, and
go for it.

>timer & bzr branch lp:python/py3k bzr-py3k & timer
Timer 1 on: 19:29:12
You have not informed bzr of your Launchpad ID, and you must do this to
write to Launchpad or access private data.  See "bzr help launchpad-login".
Branched 40811 revision(s).
Timer 1 off: 19:36:06  Elapsed: 0:06:54.44

>timer & hg clone http://code.python.org/hg/branches/py3k/ hg-py3k & timer
Timer 1 on: 19:40:15
requesting all changes
adding changesets
adding manifests
adding file changes
added 5504 changesets with 25490 changes to 4770 files
updating to branch py3k
3424 files updated, 0 files merged, 0 files removed, 0 files unresolved
Timer 1 off: 19:41:43  Elapsed: 0:01:27.32

Mercurial 1.5 minutes, Bazaar nearly 7 minutes. That's definitely
going to give an initial impression of "Bazaar is slow". And that's
with current versions of both so is it really "no longer the case"
that Bazaar is slower?

I'm curious to know if there are good reasons for the difference (that
"40811 revisions" vs "5504 changesets with 25490 changes" is
suspicious, and I guess Launchpad may still be on an old repo format)
but even if there are, it's still a reality that has to be dealt with.
It's the same with the shared repository story - no matter how much
shared repos are the "right way to go", people seem to encounter them
via the route of getting performance problems with independent
branches, and being told that a shared repo fixes this. That leaves an
impression of "slow but there are hacks to fix it" which is difficult
ground to recover.

No answers, just questions sorry...

Paul.