I was writing loosely; it's something like RFC8200+RFC4861+RFC4862, which every o/s of significance has supported for >10 years.
Regards Brian On 06-Aug-21 10:25, Toerless Eckert wrote: > I would not say "standard dual stack" because i think the requirements > can be even less IPv6 than what i consider to be standard dual stack > with our existing ANIMA RFCs: > > A) The networks data-plane can be almost (*) solely IPv4 if that is the > (enterprise, industrial,...) network stack desired in a network. > The operator of such a network does not have to know or configure any IPv6. > > (*) The only IPv6 artefact is the IPv6 link-local encap of ACP secure > channel, but that is also fully automatic and nothing the operator > would need to ever care about. And it could be optimized away through > simple follow on draft (ethernet L2 encap "optimization" for secure channel). > > B) Wrt the NOC: in pre-standard ACP implementations, for IPv4-only-network > customers, I did set up ACP-connect edge router + NAT to connect to > IPv4 only management systems / NOC. When such a NAT is included as a > feature into an ACP-edge router config it can be quite simple and elegant > (e.g.: almost fully automatic, no per-ACP-node NAT config). > > Let me know when you have an ACP edge-router implementation and i'll > be happy to whip up a IPv4/IPv6 NAT draft for it if that is what is > missing to sell the ACP solution to IPv4 only customers ;-)) > > Cheers > Toerless > (via gigantic screen) > > On Thu, Aug 05, 2021 at 10:08:29PM +1200, Brian Carpenter wrote: >> We don't require any IPv6 enablement by the NOC. All we require is standard >> dual stack on all autonomic devices, which creates link-local addresses. >> Everything else in the ACP is automatic. Why waste any effort on an IPv4 >> version? (The operator can do whatever they want in the data plane, of >> course.) >> >> Regards, >> Brian Carpenter >> (via tiny screen & keyboard) >> >> On Thu, 5 Aug 2021, 20:01 Liyizhou, <[email protected]> wrote: >> >>> Hi, >>> >>> >>> >>> RFC8994 explained the main reasons that ACP is based on IPv6 addressing >>> are simplicity and scale. >>> >>> I wonder if there was/is any interest in defining and deploying ACP in >>> IPv4-only network without requiring IPv6 enablement. >>> >>> For example, as ULA and link-local address schemes are not available in >>> ipv4 networks, whether and how to establish the ACP channel? (RFC3927 >>> defines IPv4 link-local address config, but it is not a full functional >>> substitute of ipv6 link-local address in ipv4 world in my understanding.) >>> >>> >>> >>> There might be discussions and thoughts in this aspect in earlier days in >>> the community. I would be very appreciated if anyone can point them out. >>> >>> >>> >>> Thank you, >>> >>> Yizhou >>> _______________________________________________ >>> Anima mailing list >>> [email protected] >>> https://www.ietf.org/mailman/listinfo/anima >>> > >> _______________________________________________ >> Anima mailing list >> [email protected] >> https://www.ietf.org/mailman/listinfo/anima > > _______________________________________________ Anima mailing list [email protected] https://www.ietf.org/mailman/listinfo/anima
