There seems to be consensus that trying to stop NAT in the v4 world is
futile. Good. So then we ask: "what will keep it from happening in
the v6 world?"
I postulate the following as one necessary, and perhaps sufficient,
IN ORDER TO AVOID v6 NAT: Network administrators of any home or
enterprise network need to have, at essentially zero cost, "ownership"
or control over SOME NUMBER of bits of the v6 address space,
sufficient to uniquely address each host in their network, and such
that a change in ISP or upstream topology (the higher-order address
bits) does not require reconfiguration of end-systems OR OF ANY
TRAFFIC-DISRUPTION-APPLIANCES THAT LIVE WITHIN THAT NETWORK... many of
which implement policy restrictions based on host IP addresses.
This is a refinement of my earlier comment that v6 addresses must be
both free and provider-independent in order to avoid pervasive v6 NAT.
I think that those who believe v6 DHCP and auto-conf are sufficient to
avoid the PI or local-autonomy requirement are deluding themselves,
but one cannot prove such propositions until it is too late to prevent
the undesired outcome.
What are the prospects that the condition above could be met at this
Ietf mailing list