A new article at the RIPE blog explains the current state of running IPv4 services over IPv6-only networks. The three pieces are RFC 8950 (IPv6 next-hops for IPv4 prefixes), a draft about IPv6 next-hops for IPv4 routes, and a draft for solving the IPv4 gateway problem that systems running IPv4 services in these types of networks typically have. It’s cool to see this gateway thing in such an informed piece, I’ve been wondering about that bit for a while. You might have seen different service providers solving this in different ways, with a /32 on either a regular interface or a loopback, and a strange default route pointing to just an interface (with no destination IP), or a destination IP in a prefix that the given machine doesn’t know about. Linux seems to support a couple of different ways to go about this (of course), and FreeBSD and OpenBSD each support a single, different solution AFAIK. I don’t believe illumos even has a way to solve this problem at the moment, but I might have that one wrong. Anyway, do have a look at the article: https://labs.ripe.net/author/remco-van-mook/a-farewell-to-arps-ipv4-service-on-ipv6-only-networks/ #IPv6 #IPv6mostly #fuckdualstack #networking