A recent hands-on experiment with RFC8950 shows how IPv4 can be routed over an IPv6-only network without tunnels. The standard extends Multi-Protocol BGP so that IPv4 routes can carry IPv6 next-hops. Only the control plane changes: IPv6 handles route exchange and neighbor discovery through NDP, while IPv4 packets still travel as native IPv4 over Ethernet frames.
The author highlights several practical uses. A large network could run IPv6 only in its core while leaving IPv4 at the edges, requiring no changes from customers. An IP transit backbone could exchange IPv4 routes with its customers, peers and upstreams over IPv6 sessions, eliminating much of the IPv4 interface, BGP and firewall configuration. Internet exchange points could also offer IPv4 services over a single IPv6-only LAN and route server, much as 464XLAT did for mobile networks.
Support is already broad. Juniper Junos, Cisco IOS XR, Nokia SR OS, MikroTik, Huawei, BIRD and FRR all implement the feature, so recent software versions are likely to work.
The demo lab, built with Containerlab and open-source FRR, links two customer routers through three provider routers. The core runs IPv6 only, while the second customer uses an IPv4-only enterprise link. The two customers can still reach each other across the IPv6 core without any tunnel.
Source: Hacker News · Summarized by HeadlinesBriefing