Agreed, ND proxy shouldn't be normalized. But the hack is the same whether it's automated or not; the difference is whether users get working IPv6 or give up because it's too hard to set up. The latter is worse: if users give up on IPv6, ISPs may conclude they needn't provide IPv6 at all, let alone delegated prefixes. I'm thinking of logging a warning whenever the fallback kicks in, suggesting users ask their ISP for PD.
You're right about RFC 7278, I misread it. It only describes a routed setup, moving the /64 to the LAN. So on point-to-point WANs I'll do exactly that, and use ND proxy only when the /64 is on-link on a broadcast WAN.
Things are improving, though. In Japan, NTT's newer 10G service delegates a /56 by default; only the older 1G service without a phone subscription is stuck with a single /64.
You're right about RFC 7278, I misread it. It only describes a routed setup, moving the /64 to the LAN. So on point-to-point WANs I'll do exactly that, and use ND proxy only when the /64 is on-link on a broadcast WAN.
Things are improving, though. In Japan, NTT's newer 10G service delegates a /56 by default; only the older 1G service without a phone subscription is stuck with a single /64.