Hi Remco, On Tue, Oct 21, 2025 at 10:24 PM Remco van Mook <[email protected]> wrote: > I'd like to draw your attention to a draft I submitted last night - > https://datatracker.ietf.org/doc/draft-ipv6-resolved-gateway/ . > The short version is to get a standardised approach for hosts to send IPv4 > traffic to the link layer address of their IPv6 default gateway, bypassing > the need for v4 subnetting and link layer discovery (ARP) to send IPv4 > traffic upstream. > This aligns fully with the current v4-via-v6 efforts that are currently > ongoing inside of the IETF, and extends it to the host level. The approach is > to assign an address from the IPv4 special purpose registry to this end - i.e > if this IP address is found as default gateway, the host should use IPv6 > NDP/RA to resolve the link layer next hop. This can be made fully backward > compatible by having the gateway also send ARP replies for this IP if needed. > > I presented on this during RIPE91, currently in Bucharest. Due to > overwhelmingly positive response (and a midnight deadline that I was unaware > of until then) to just submit the I-D in its current rough form.
Thank you for putting this together! I think it's a very cool idea! Obviously, nobody expects -00 to be ready for the WGLC, the purpose of -00 is to discuss if it's smth the WG is interested in, so I'm not going to pick up on the text. However I'd like to highlight some area the draft should discuss in more details: 1. A host might have multiple IPv6 default gateway, even if the host has a single network interface, especially in a multi-router/multi-prefix environment (see RFC8028). Therefore the mechanism you are proposing might be a bit more complex than "take the nexthop for ::/0, find the MAC in the Neighbor cache, put it into the ARP table" - the default address selection might get involved. Again, it's not an objection, just an area to think about. 2. It would be nice to describe an incremental deployment scenario - what would a "legacy" host do when it gets an IPv4 address with /32 mask and 192.0.0.11/32 as a default gateway? Would be interesting to see how existing OSes behave (a hackathon?). 3. Operational considerations: what are the implications for the Ops team? How can they use their favorite tools like traceroute and ping? I'll try to find time and send you a GitHub PR for some of the points above. > The presentation can be found here: > https://ripe91.ripe.net/programme/meeting-plan/sessions/10/Y8YVLE/ > > I would very much like to hear your opinion - I've asked the chair for an > opportunity to present on this at the Montreal meeting. I hope the chairs would be able to fit your talk into the agenda! -- Cheers, Jen Linkova _______________________________________________ Int-area mailing list -- [email protected] To unsubscribe send an email to [email protected]
