[darcs-users] performance; clone?

[Dan Egnor]

Darcs is very interesting.

We noticed that if you fetch the darcs repository, undo a few dozen
patches, and then pulled the repository again, it takes a very long 
time (killed it after running for 45 minutes) -- much longer than it 
took to get the repository in the first place.  Is this known to be slow?

Given the storage format, it actually surprises me that anything is
fast; I can edit a file and run "whatsnew" and it doesn't take a long
time.  I wonder how that works.

Do you believe that all operations can be made relatively fast with
the addition of elbow grease (perhaps by maintaining a cached index
file, updated or reconstructed as needed)?

I was also wondering how difficult it would be to support a "clone"
mode of operation, where you could "get" or "pull" from a local
repository (rather than a remote URL) and instead of copying the
patch files it would create a (symbolic? hard?) link.  That way you
could create local branches relatively quickly on a shared system,
and the filesystem buffer cache would be shared between the original
and the new copy.  (Patch files are never modified after being recorded,
right?)

Dan