I use CVS so much (and use the actual ,v histories so much) that it 
justifies keeping a mirrored tree of the cvsroot locally in my 
network here.

I do this using rsync against anoncvs.netbsd.org.

Often there are complaints that rsync eats up huge amounts of memory 
dealing with large file lists and doing its thing, so I thought I'd 
try to go easier on the rsync server by just rsync'ing smaller 
portions of it at a time, with sleeps in between.

I have a small script which retrieves the current list of 
directories under the rsync path by:

rsync rsync://anoncvs.netbsd.org/cvsroot/

... and then under each of those paths:

rsync rsync://anoncvs.netbsd.org/cvsroot/"$i"/

... and then I finally rsync those components one at a time, waiting 
one second between iterations.

Can anyone see anything wrong with this kind of usage? I realise 
that, should a single file appear at the /cvsroot/X level this 
won't work; however, it seems to me that averaging out usage over a 
longer period of time is friendlier to the server resources.

Thanks in advance for your comments,
Marc Tooley

P.S. For posterity, I humbly request that you direct your responses 
to the list itself, as in the Reply-To: header.