In the near term, we should recommend more liberal address assignment policies so that multiple prefixes and renumbering are not needed
The problem with this is that multiple addresses were adopted as the way to do large-scale multi-homing (i.e. having a lot of multi-homed sites) because it was the only approach that seemed technically feasible within the existing architecture (both routing, and the various namespaces).
if it seemed technically feasible, perhaps it was because a problem one hasn't tried to solve yet (and which is "somebody else's problem") seems more feasible than a problem one is familiar with.
It may well be true that in practise, having several addresses (in the sense of "names that identify both location and identity") is unworkable. If so, that inevitably means that scalable multi-homing is not practical with IPv6 - and it also inevitably means that the multi6 effort ought to be abandoned.
we haven't determined that the situation is unworkable, but neither have we found a way to make it workable. also, there is room for compromise. It appears to me that renumbering and multihoming can be made manageable with some additional protocol support (HIP being potentially one of the pieces, but only one of them) _and_ with some constraints on the use of multiple address prefixes. (at present there are many more pressures for hosts to have multiple addresses than just v6 renumbering and multihoming)
_______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf