At the moment, identifying the remote repository format takes (for whatever
reason) a large chunk of the time of push/pull according to status output of darcs
So i propose a (maybe on personal configuration basis) change that would cache
the remote repository format and only re-identify it when a problem occurs.

I agree this could be nice, as long as logic of "deciding when a problem occurs"
that should trigger a re-cache is clear enough.

This is probably only apparent when accessing a repository over ssh? In
this case, the slowness only appears if darcs2 is not present on the remote
server, which is something that should become less common with time.  So I
don't think this is worth fixing.
-- 
David Roundy
Department of Physics
Oregon State University