Bob,
As IPv6 build IIDs out of MAC addresses, should the IETF make rules on how MAC addresses are assigned? From what I remember there is an agreement between the RIRs and the IETF on who does what, right? Making the claim above implies changing a lot more than IETF specs.The policy has to be consistent with the technical constraints. The IETF needs to define what the technical requirements (i.e., what works, waht doesn't work, what the limitations are, scaling properties, etc.) in order for the RIR to create the policy that is practical.I think the answer is that the RIRs can do allocations based on the technical requirements given to them by the IETF.Hmm. This depends on what you mean. Policy is today decided by the membership of the RIRs. I think that RIRs should do allocations according to the policy set by the RIRs members. The IETF should set technical standards for the protocols using those addresses or making up those addresses. If those protocols require limitations in policy, that should be declared in the protocol specification. I think you will find it hard to get the RIR membership to agree on having policy set outside the RIR membership.
You can do this in multiple ways. You could (playing the devils advocate) argue that the IETF should create technical specifications that can handle the policy of the RIRs. After all, to a large extent the RIR membership is most likely a better representation of "end-users" (for some definition of users) than the IETF is. A better way to move forward is most likely to have the IETF _cooperate_ with the RIRs on a working policy.Agreed. But the policy needs to consistent with the technical requirements. For example there are technical limitations of routing protocols that relate to the number of routes and interconnections between ISPs. If the RIRs were to adopt a policy that did not take these technical requirements it would result in an Internet that didn't work.If you want to influence the allocation policy, go the RIRs meetings.