Dave, >> my view is entirely different. >> the 64 bit boundary is a suggested policy and not normative. > > That's not true. RFC 4291 is the addressing architecture and normatively > states: > > For all unicast addresses, except those that start with the binary > value 000, Interface IDs are required to be 64 bits long and to be > constructed in Modified EUI-64 format.
right, no RFC2119 language there. apart from that, implementations support and people are using prefixes longer than /64 today. so we can either choose to put our head in the sand or adjust specifications to reality. > Furthermore 129 addresses in every subnet are reserved. See RFC 2526. > > Within each subnet, the highest 128 interface identifier values are > reserved for assignment as subnet anycast addresses. > > These are assigned by IANA > (http://www.iana.org/assignments/ipv6-anycast-addresses) > > So you cannot have any unicast addresses in any subnet with a prefix > longer than /120. However, the point is, you don't need to. I think the purpose of this draft is to say that you can. but you cannot use reserved anycast addresses on these links, nor can you make assumptions about the u/g bits. > You just use the off-link model. > >> IPv6 is not classful. >> two routers on a /127 are on-link to each other. two /128s would make them >> off-link. > > Be careful not to confuse the terminology. The definition of "on-link" is in > RFC 4861. what definition do you use? if you want to define two nodes as off-link to each other you might use two /128s. if I use a /127 then this bullet of 4861's on-link definition applies. i.e these two nodes are directly connected, as they have addresses out of the same /127. - it is covered by one of the link's prefixes (e.g., as indicated by the on-link flag in the Prefix Information option), or (it appears to me that people coming at this from a routing perspective have quite different models in mind than people coming at it from a host perspective...) >> a off-link node cannot make any assumption on the use of subnet-router >> anycast addresses, reserved anycast space or the validity of u/g bits. > > That's not true either. The RFCs quoted above have no such restriction > on what off-link nodes can assume. my point is that an implementation cannot assume that setting the last 64 bits to zero will give you a valid subnet-router anycast address. nor that taking a random 64 bit prefix and using the reserved anycast address will work nor that you can glean anything certain from u/g bits. cheers, Ole -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------