Alexandru Petrescu [EMAIL PROTECTED] wote: > >>> In that case, there will be old cards which can't support 33:33 > >>> MAC addresses. Perhaps it is well to note that these cards > >>> won't be able to run IPv6. > >> > >> Ok, so that's an option. Another is to just say that when 33 > >> multicast is not available for broadcast, just use ff broadcast. > > > > > > So the obvious question is... how do we know if a receiver on the > > link cannot use 33 multicast? I dont know of a way for the IP stack > > to find this out. > > Right. The problem is I don't know about about a way for the IP stack > to find out that the receiver _does_ use 33 multicast. It simply > assumes it. The RFC assumes it.
I thought the RFC specifies the use of 33:33 multicast rather than assumes it? > > > But assuming that we can and do detect such a reciever then all > > multicast RA/ND messages from a router would then have to use ff > > broadcast? > > No no, only the RA and NA messages that are intended to be > broadcasted, > and broadcasted to the complete set of members, not just some group > address derived from a specific MAC address. > > For example, RAs can be sent to either 33:33::1 or to specific MAC > addresses too. Those that can be sent to 33:33::1 should be > able to be > sent to all-ff too. It's just a change in notation. There > is no change > in scalability or power consumption. Ok, this is a superfluous argument but if the router *could* detect a non 33:33 capable node on it's link it would have to either 1) send all multicasted RA/ND messages to the all ff address (so that the non 33:33 node can see them) or 2) unicast the multicasted RA/ND messages in duplicate to the non 33:33 node. > > > Perhaps it would be simpler just to add some text saying that > > Ethernet NICs that do not support 33 multicast may not (cannot?) be > > able to support v6? > > "cannot" is too strict. IPv6 is a big thing. Only a certain form of > easy stateless autoconf is not supported if the 33 multicast is not > supported. If a non-33 host sends a RS to a pre-configured > router then > that host does receive a directed RA. > I agree "cannot" is too strong. However, the normal (according to RFC 2461) RA response to a RS is sent to the all nodes multicast address (FF02::1 i.e. 33:33::1) rather than the specific MAC address of the soliciting node. Although most implementations may allow you to change this to a unicast response, I'm fairly sure the default settings will be for a multicast response. Martin -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------