bzr wins a performance test

[Theodore Tso]

On Wed, Jan 23, 2008 at 11:15:45AM -0500, Jonathan S. Shapiro wrote:
> On Wed, 2008-01-23 at 07:52 -0800, Gé Weijers wrote:
> 
> > As far as I'm concerned this is fairly irrelevant, given that on most
> > projects the size of your checked out files dwarfs the repository
> > size. Speed and ease of use should win every time.
> 
> This is completely correct, but the conclusions about file size need to
> be made with caution. Many files are very small, and the block tail
> effects can become quite large. This is especially an issue with
> directories, and can also be an issue with repository metadata.

All true.  In addition, the comparison involved only one version of
the source tree.  You really want to import the same repository with a
large number of changesets into multiple SCM's repositories to see how
much space a real-life repository with a substantial amount of history
takes up in each SCM.

In addition, there usually tuning knobs that control tradeoffs between
compression efficiency and time spent compressing, and then, of
course, there is the question of how long does it take to do various
operations on a compressed repository.  Often it is faster because you
have less data to read in from disk, but not always.

So it's very hard to make a comparison, and basing it just on space is
clearly insufficient.

						- Ted