[profmonocle]

> There must be new IPv8 with backward compatibility with ipv4

How would you accomplish this? How could an IPv4-only system send a packet to an IPv8 address? IPv4 packets only have 32 bits for the destination and source IP addresses, there's no way to fit the address for any newer protocol with longer addresses in there.

[octoberfranklin]

Mandatory NAT64 support in every dual-stack router.

Concretely: if you're a router and you accept IPv6 packets, you must offer NAT64 services on any interface from which you accept IPv4 packets. The upgrade ripples outward from the core routers (default-free zone) in two waves. As soon as all of a router's peers support IPv6, that router can drop NAT64 and IPv4. No flag days. All coordination is between a pair of peers with a direct link.

There were IPng proposals that had this requirement (using different terms, of course). IPv6 wasted more than a decade fighting against NAT64, tooth and nail. It wasn't until they finally caved and standardized NAT64 that people took IPv6 seriously -- but by then it was too late to make it a mandatory feature of dual-stack routers and there was already hardware shipping with IPv6. The clean upgrade path was lost forever.

Next time anybody complains about the IPv6 transition taking so long, you know who to blame.

[kortilla]

Funny that you think NAT64 is relevant. Every successful deployment I’ve worked with in the US is just dual stack. Native v6 networks with nat64 came after the big content providers got ipv6 server support for the dual stackers.

NAT64 has worse scaling than CGNAT (requires more state and moving pieces) so it was never a solution to drive adoption.

[octoberfranklin]

Nearly all IPv6 traffic is cell phones. And they are all behind NATs.

[kortilla]

You got confused. If it’s ipv6 traffic it’s not behind 6to4 nat. Any stats about ipv6 on the internet are by definition not included stuff that got hit with 6to4.