[Ospf-wireless-design] Two clarifications about OSPF-MDR

I want to make a couple of clarifications/corrections
to help understand (or avoid misunderstanding) OSPF-MDR.

Phil and I assumed uniconnected adjacencies in our simulations,
since the MPR approach does not ensure biconnected redundancy
(backup MPRs could be used for that).  I want to clarify that
we don't have to decide between uniconnected and biconnected
adjacencies, since they are designed to be interoperable.
I.e., the protocol still works if some routers choose
uniconnected and others choose biconnected.

This is why the condition for keeping an existing adjacency
is the same for both uniconnected and biconnected adjacencies:
an existing adjacency is kept as long as either endpoint is
an MDR or Backup MDR.  Thus, if one endpoint is using
uniconnected adjacencies and the other endpoint is using
biconnected adjacencies, the two nodes agree when the
adjacency should be eliminated.

Biconnected adjacencies provide a little more robustness,
but uniconnected adjacencies should be useful for achieving
scalability to a large number of nodes (perhaps 300).

I also want to modify a comment I made regarding the Wait Timer.
A router that starts running OSPF-MDR should actually wait
HelloInterval * 2HopRefresh seconds before running the MDR selection
algorithm, since a full state Hello is sent every 2HopRefresh Hellos.
If diff Hellos are not used, then 2HopRefresh = 1, so the waiting
period should be HelloInterval.  If diff Hellos are used, then
HelloInterval * 2HopRefresh is typically equal to RouterDeadInterval
(with HelloInterval = 2 and 2HopRefresh = 3).

Richard

_______________________________________________
Ospf-wireless-design mailing list
Ospf-wireless-design@lists.ietf.org
https://www1.ietf.org/mailman/listinfo/ospf-wireless-design