> "A multi-link subnet model should be avoided. IETF working groups > using, or considering using, multi-link subnets today should > investigate moving to one of the other models."
Sure. We knew about the problems and did investigate whether there were other models. RFC 4903 was good advice, indeed. But, in the end, divorcing the ephemeral link concept from the more permanent subnet was the right way to run constrained node networks. You will notice that many arguments in RFC 4903 are about limits of ND classic. So we fixed up ND for our purposes in RFC 6775. (Some of that may still flow back to "big" networks in the form of Efficient ND.) The other thing that we threw overboard was the fiction that you can have multicast working in every subnet, at the same low expense of a simple unicast packet. We gave up on that and now have MPL (draft-ietf-roll-trickle-mcast) as a (decidedly expensive) alternative, with some changes in the service model. (We may still get some more traditional IP multicast back by using variants of the BIER model, see e.g., draft-bergmann-bier-ccast. But the parity of multicast with unicast is not going to be restored.) In summary, there were some changes, and some compromises, in the architecture. Still, the result is much more livable than the classic MANET model combined with the Mobile-IP model. Grüße, Carsten _______________________________________________ Int-area mailing list [email protected] https://www.ietf.org/mailman/listinfo/int-area
