On Wed, 14 Jul 2010 13:16:24 +0200 (CEST) [email protected] wrote: > > For point-to-point links, which can do multicast trivially, there is no > > excuse > > for not doing full ND (unless it is a link between two routers that actually > > use hellos in the routing protocol to determine whether the other side is > > alive). > > However, it would seem that several of the major vendors (e.g. Cisco, > Juniper) have interpreted this differently, and chosen not to perform > ND on point-to-point links. >
Well, if my understanding of the RFC is correct, and these implementations (and others I'm aware of) are claiming compliance with the ND RFC, then they're suffering from implementation bugs. That's fine, bugs are inevitable. I think a feasible explanation could be that the model of IPv4's operation over point-to-point links has been copied when implementing IPv6, even though one of the key differences between IPv6 and IPv4 is the shifting of address resolution into ICMPv6. On a PPP link, IPCP conveyed information about complete IPv4 addresses at the other end of the point-to-point link. That functionality been removed from IPV6CP - it only negotiates/conveys IIDs. So it also seems that the IPv6 over PPP authors were assuming that something else would take over that role, most likely ND NS/NA. Routing protocols would only partially fill that role, because they won't be operating on PPP links between e.g. residential ADSL CPE and the upstream customer aggregation router. That will be an exceedingly common scenario, and one where the /127 mitigation can't be applied because of it's manual configuration requirement. -------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
