On 15/10/2014 08:31, Ted Lemon wrote: > On Oct 14, 2014, at 2:19 PM, James Woodyatt <j...@nestlabs.com> wrote: >> On the topic of the original question, if I were to editorialize here, then >> I would want to see something like this: > > I get that you have an opinion on this, but you haven't actually stated any > argument to support what you think we should do. And there are some > implications in what you are saying that I don't think are necessary. > >> A) An autonomously generated ULA prefix SHOULD be advertised when no other >> delegated prefix is valid. > > OK, although underspecified. > >> B) Whenever there is any valid delegated prefix, advertisements for an >> existing autonomously generated ULA prefix MUST be deprecated, i.e. updated >> with preferred lifetime of zero. > > Why? What problem does this solve? Given that it's going to mean > additional work, there should be some benefit to doing it.
At a stroke this would destroy the main advantage of ULAs - namely, invariant addresses for internal traffic. IPv6 assumes multiple simultaneous addresses; there is no reason whatever to artificially prevent use of ULAs alongside GUAs. Brian >> C) A deprecated autonomously generated ULA prefix MUST be withdrawn when it >> expires, i.e. its valid time reaches zero. > > Okay, given that a prefix expires, it should be withdrawn, whether it's a ULA > or a GUA. > >> D) Whenever there is no longer any valid delegated prefix, advertisements >> for a previously deprecated autonomously generated ULA prefix MUST be >> updated with non-zero preferred lifetime. > > OK, but seems like unnecessary work. You're essentially recapitulating the > brokenness of IPv4 zeroconf. > >> The idea here is to make sure IPv6 applications can generally rely on home >> network interior routers to forward traffic among the multiple links in the >> home, regardless of whether any first-mile Internet services are >> provisioned, configured and operational, i.e. there shall always be at least >> one preferred global scope network prefix, and there shall be an >> autonomously generated local prefix available as a last resort whenever >> there are no valid delegated prefixes. > > This is where I am just completely puzzled. We talked about this > previously. I thought the idea was that the homenet ULA should converge: > that there should only be one, ultimately, and that when there are two, > routing should still work. You are stating this as if the ULAs are > per-subnet of a homenet, and that routing across homenet routers using ULAs > isn't supported. > > If you really think that's how this should work, I can see why you want to > deprecate them. But that's not how they should work. > > _______________________________________________ > homenet mailing list > homenet@ietf.org > https://www.ietf.org/mailman/listinfo/homenet > _______________________________________________ homenet mailing list homenet@ietf.org https://www.ietf.org/mailman/listinfo/homenet
