On 10/06/2017 02:48, Michael Richardson wrote: > > Brian E Carpenter <[email protected]> wrote: > > So this is why I've been asking the ACP authors to be precise about the > > service they will offer to GRASP, over many months. They tell me it > > will be a VRF instance providing a reasonably small number of (virtual) > > interfaces to GRASP. GRASP will send LL multicasts to each of those > > interfaces, and the VRF instance has to do the right thing, i.e. > replicate > > those multicasts as needed. But that's layer 3, it just has LL scope. > > > Nothing new here that I can see. > > Since all the links will be PPP links, there will be no "replication".
I think that's our choice to make. Please see: https://mailarchive.ietf.org/arch/msg/anima/-vOvFVQr6zcognXJxwKKB9_t5-Q GRASP discovery and flood, by their nature, must be received by all GRASP neighbours. How it happens isn't actually very important - if there was such a thing as secure link-local multicast, we'd use it. But if not, *somebody* is going to replicate the packets. If the VRF doesn't do it, GRASP itself will have to do it. That doesn't seem optimal to me. I would like the ACP VRF to emulate link-local multicast to all nodes listed in the adjacency table as neighbours on the same physical interface. If it doesn't, then GRASP will. IMHO we need a precise statement of the service the ACP will provide at the API it presents to GRASP. That can't be left as an implementation detail. Brian _______________________________________________ Anima mailing list [email protected] https://www.ietf.org/mailman/listinfo/anima
