Benchmark: Git 1.6.3.3, Hg 1.3.1, Bzr 1.17

[Stephen J. Turnbull]

Jari Aalto writes:

 > So for remote cloning
 > 
 >         bzr branch REMOTE HERE      # Bad

No, this decision is determined by the workflow.  Disconnected
workflows *require* a full repo clone.  Distributed workflows are
mostly much happier with a full clone (but there are exceptions).
On the other hand, full clones are inefficient and maybe error-prone
for centralized workflows.

 > But
 > 
 >         bzr init-repo --2a
 >         bzr checkout REMOTE HERE.trunk
 >         bzr branch HERE.trunk HERE.working
 > 
 > Is this more efficient 'clone' ?

IIRC, the checkout isn't a clone at all, and IIRC you can't branch
from a checkout anyway.  Even if you can, I find the semantics I would
associate with that to be very confusing.

BTW, if you're going to get into checkouts and stacked branches for
bzr, you should look into comparing to git shallow clones (git clone
--depth=N ...).

But not really ... the point is that bzr is *way* too complex for this
kind of benchmarking to make a lot of sense, and it gets worse when
you try to trade off options against the other VCSes .  Some of the
basic operations are now approximately in the ballpark, and that's the
best you're going to be able to say unless you want to focus on a very
specific workflow.