[savi] Need Binding Distribution Protocol in SAVI?

[Christian Vogt <christian.vogt at ericsson.com>]

Dear SAVI Folks -

We have recently discussed the need for, and the possible properties of,
a binding distribution protocol for SAVI deployment on multi-switch
links.  I would like to take this discussion up again.

The purpose of a binding distribution protocol is to enable multiple
SAVI devices on the same link to synchronize their bindings.  There are
three main approaches to design such a protocol:

(1) Let SAVI devices synchronize by listening to existing protocols
    (such as DAD, NUD, ARP).

(2) Allow SAVI devices to actively participate in existing protocols.

(3) Design a new protocol for SAVI device synchronization.

In order to move forward in this matter, we should get a better
understanding of the benefit/cost ratio of possible solutions.
Obviously, we will have to make a tradeoff:  For maximum spoof
protection, a new binding distribution protocol may prove to be
indispensable.  But perhaps we can gain simplicity by making small
concessions in terms of spoof protection.

Perhaps we should start by evaluating how reliably SAVI devices could
synchronize with option (1), the simplest of all options.  How would
option (1) perform for DAD, how for Stateless Autoconfiguration?

Comments appreciated.

- Christian