On 03/06/2009 11:09, Ingrid Wijte wrote:
IPv6 Provider Independent (PI) Assignments for LIRs
This policy proposal has merit, and is really a variety of the inherent aggregation "problem" which was "solved" in ipv6: namely that with ipv4, providers got smallish blocks of space which they could chop and dice in all sorts of flexible ways under the category of traffic engineering (because as we all know, deaggregation is bad and should be stamped out). With IPv6, the underlying assumption has been that /32 "should be enough for anyone". Indeed, for most people, it is - if you want to stick all your traffic engineering eggs in the one basket. But now that we're actually deploying IPv6 a little, it's becoming clearer that this is an operationally naive assumption to make. Turning it around slightly, the problem is an extension of the multihoming problem. We have no functional technical solution for this at the moment, other than to use the ipv4 bodge: namely multiple prefixes, each announced with an independent routing policy. So, really, the issue you're trying to solve here is probably not whether LIRs should be entitled to PI assignments (which is perhaps an interesting digression), but really how can LIRs engage in meaningful traffic engineering in the ipv6 world, and whether this should interact with address policy management. I sense a collision of policy and operations here. Nick