> On 2Oct2017, at 17:21, Toerless Eckert <tte@xxxxxxxxx> wrote: > > Of course, this attempt to get best of both worlds would not work for extensions > to existing IPv4 only protocols like OSPFv2 or ARP or the like, which is why i am > asking A) - to understand how much of a problem those protocols are. With respect to OSPF, my understanding is: - OSPFv2 can only be carried by IPv4 and only carries IPv4 prefixes. - OSPFv2 is extraordinarily widely deployed in enterprise IP networks (and residential networks). I very much doubt that many those enterprise deployments will drop OSPFv2 anytime soon. - OSPFv3 now has clear specifications (but RFC-7949 is quite recent) to enable OSPFv3 to be carried by IPv4 and by IPv6, BUT I am unclear how widespread support for that extension is in shipping commercial IP routers at present. - OSPFv3’s Address Family extension (RFC-5838) permits any OSPFv3 instance to carry both IPv4 routing information and IPv6 routing information. I am unclear how widespread support for this extension is in shipping commercial IP routers at present. - I believe some OSPFv2 extensions are being undertaken — e.g., “segment routing” (sic) extensions that are equivalent to the corresponding OSPFv3 “segment routing” extensions. - Given the huge deployed base of OSPFv2, it makes sense to have equivalent extensions in both OSPFv2 and OSPFv3 for some while yet, IMHO. One of the Routing ADs or OSPF WG Chairs might be able to clarify/confirm the beliefs outlined above. With respect to ARP, I do not know of any current work, but if some new security situation arose, then the IETF ought to be able to make appropriate changes and/or extensions to the ARP specifications. Cheers, Ran