partitioning and referrals

[Dan Shriver <dshriver@sharemedia.com>]

In doing some more reading I see that I used the wrong
terminology in describing my problem.

I'm trying to make a distributed directory, "Understanding and
Deploying LDAP" refers to what I'm trying to do as
"partitioning"

			A
		      /   \
		     B     C

Simple example: A is a master server, containing the top part of
my directory.  Other slave servers (B & C) each contain distinct
subtrees (they are not found in A, only a reference to them
exists in A, nor are they found on the other slave servers).

My understanding is that all changes to the directory go through
the master.  So after setting up the master & slave with the
proper referrals, I tell 'A' add these entries.  'A' looks in
its tree and notices that all the entries are actually in a
subtree deligated to 'B', the entries are then sent to 'B' with
the proper binddn and added.

I see that replication and referrals are different (and
partitioning is also different) where I am confused is that the
referrals are all 'A' has to determine where to add entries
(Likewise if 'A' contains the entries in 'B' and 'C', but 'B' &
'C' are distinct vs each other the situation is similar, 'A'
needs to know which machine to replicate which entries to and
the only piece of info I've seen is referrals) or is there
another object type that I should be using?