Brian, (a) The fact that IEEE MAC addresses are recycled by some vendors is well understood. Please be sure I don't ignore it.
(b) I am also well convinced that u=1 isn't a "mathematical guarantee" of uniqueness. To me, what is important is agreeing that the 4rd IID range makes sense. (Despite my understanding that the point I had made below isn't contradictory with (a) and (b) above, I see no point in arguing more about it.) Thanks, RD Le 2013-02-05 à 15:36, Brian E Carpenter <[email protected]> a écrit : > > On 05/02/2013 14:11, Rémi Després wrote: >> Le 2013-02-05 à 15:01, [email protected] a écrit : >> >>>> With these specifications, it is impossible on a SLAAC link that two hosts >>>> that have universal-scope addresses (necessarily different if they >>>> communicate at the MAC layer), would have conflicting IIDs at the IP layer >>>> if they derive them from MAC addresses. >>> Impossible if the MAC addresses are indeed unique. That is correct for >>> *almost* all cases. But not 100%. >> >> They are different 100% of cases at the IP layer if they are different at >> the MAC layer (which is needed for communication at this layer). >> OK? > > Not if the same MAC address appears on two different subnets in the same > site. > This would work from a conventional IPv6 point of view, but it would > immediately demonstrate that the u bit does not mathematically guarantee > uniqueness. > > There has been discussion recently in MIF of scenarios where the same > IID would appear intentionally on different interfaces of the same device. > > It is not the process of generating IIDs that is a problem - that is > well defined. It is assuming after they have been generated that the > u/g bits have meaning that is problematic. (Which is also why the > solution of a reserved range makes sense for 4rd.) > > Brian > -------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
