Re: [OSPF] Questions regarding draft-ietf-ospf-manet-or-00.txt

[Richard Ogier <ogier at earthlink.net>]

Abhay,

I have a few more questions regarding the OSPF-OR draft.
The first bullet of Section 3.3.3 (Terminology) states:
"Note that all neighbor references in this document are
assumed to be FULL neighbors."

Is this really true?  If so, then it would imply that all overlapping
relays (ORs), including non-active ORs, must be Full neighbors.
In particular, an adjacent neighbor in state Exchange cannot be a
non-active OR (even though OSPF performs flooding along all
adjacencies, not just Full adjacencies).

Section 3.3.8 (Flooding and Relay Decisions) states:
"Upon receiving an LSA from an adjacent speaker, a node makes
flooding decisions based on the following algorithm."

The above is consistent with OSPF flooding.  However, if only Full
neighbors can be ORs, then a received LSA will never be relayed
unless the LSA was received from a Full neighbor, which does
not seem right.

The GTNETS code for OSPF-OR developed by Boeing, and used for their
Milcom'06 paper, actually allows an received LSA to be relayed even
if it was received from a non-adjacent neighbor.  This implicitly
allows any bidirectional neighbor to be a non-active OR.

So there are three possibilities for (non-active) ORs:
1. An OR must be a bidirectional neighbor.
2. An OR must be an adjacent neighbor (state Exchange or above).
3. An OR must be a Full neighbor.

When smart peering is used (for adjacency reduction), the first
choice might have the drawback that there are too many non-active
(backup) ORs, which can affect scalability (as shown in simulations
of large networks).  Choice 2 would limit ORs to the reduced
adjacency graph, which would result in far fewer non-active ORs.
Although the current spec implies that choice 3 is used, as I
mentioned above this does not seem like the best choice.

If the answer is either 1 or 2, then the draft should be modified to
change the definition of an OR.  For example, if the answer is 2,
then the definition can be changed to:

"An adjacent neighbor is considered an overlapping relay for a
speaker if it has an adjacent neighbor that is not an adjacent
neighbor of the speaker itself."

Richard
www.manet-routing.org

_______________________________________________
OSPF mailing list
OSPF at ietf.org
https://www.ietf.org/mailman/listinfo/ospf