On Jul 20, 2009, at 11:20 AM, Fred Baker wrote:
On Jul 20, 2009, at 10:34 AM, james woodyatt wrote:On Jul 20, 2009, at 09:07, Iljitsch van Beijnum wrote:Most important issue: "The CPE Router may also support prefix sub-delegation."We should make this a MUST. I don't want to end up in the situation where I get a crappy DSL IPv6 CPE from my ISP and I can't use my own CPE because there is no subdelegation.Careful with that axe... if we're going to require prefix sub- delegation in best-practice CPE routers, then I contend we absolutely need to specify how that's done with zero configuration. Are we ready to do that yet? I don't believe we are. Very happy to be proven wrong about that.Personally, I am uncomfortable with the topic's treatment in this document.draft-donley-ipv6-cpe-rtr-use-cases-and-reqs-00.txt and draft-ietf- v6ops-ipv6-cpe-router-00.txt both comment on sub-delegation; the latter tries to specify an algorithm for it. In a document of this type, to my mind, the question is "what RFC MUST/SHOULD/MAY a CPE Router implement", not "what algorithm...". That also reflects the discussion at IETF 74, which recommended dividing the CPE Router Recommendations document into a "2009 recommendations" that would be a working group document and listed RFCs, and a "2010 recommendations" that was an individual submission and requested RFCs on various topics to use in future recommendations. I would be far happier with specifying that
reword...
a) a CPE router with one external and multiple internal interfaces that is delegated a prefix shorter than /64 from its upstream SHOULD assign specific /64s from that to its non-upstream interfaces by an algorithm internal to itself, and
b) It would be Really Nice if the 6man working group designed a multi-router sub-delegation algorithm for cases in which a small network with multiple routers needs to do so within itself.
By the way, there are many words of caution that apply to auto-sub- delegation of prefixes. In general, one wants them to aggregate in a nice way without all the issues of renumbering. That means in part that there must be some way to recognize when it is inappropriate to use the algorithm, such as saying that it is appropriate to networks using PA prefixes and in which internal aggregation is not in use. If my upstream gives me a /60, there isn't a lot to discuss. If the upstream gives a /56 or a /48 and the site in question has multiple campuses, auto-sub-delegation is probably pretty difficult to get right.
To me, anything beyond that is not a document making appropriate "recommendations" to vendors regarding their products, or to the likes of CableLabs or BBF regarding the products they deploy.I don't see an RFC or Internet Draft in any working group that tries to specify sub-delegation for SOHO/SMB networks. Fred Templin has three documents that look at the topic, but to my knowledge they have not been picked up by any working group.