how to simplify mutiple replication agreements

[Dan Shriver <dshriver@sharemedia.com>]

I have a complex LDAP archetecture that I'd like to simplify.

		R
	       / \
	      N   M
	     / \ / \
	    A...ZA'.Z'

Basically everything (all the data) lives in R.  N and M are
subsets (not a simple subtree of R but a collection of subtrees
from R).  A...Z and A'...Z' are logical subtrees of N and M
(think of them as geographic regions) BUT they aren't
necessarily equal (A' might not equal A (for the geographic
region example N might cover 5 cities in a given county (but M
might cover all the cities in that county))).  This bothers me
because it means that I have to subdivide the true regions under
R and replicate huge numbers of small subtrees (in order to
allow for A not always equalling A').  Basically it looks like
I'd need to do 10,000's of replication agreements (so I'd be
running 10,000's of slurpd's...).  How can I simplify this so
that I only have a few slurpd's running?  Has anyone done stuff
like this before with openLDAP?