On Feb 23, 2013, at 14:24 , Fred Baker (fred) <f...@cisco.com> wrote: > > One: the v6ops result reflects the operational result in the ARIN community: > operators there would like to be able to allocate /56 prefixes to smaller > customers and /48s to larger ones. If you want castigate someone, castigate > them.
I'm unable to participate in ARIN or other RIR discussions, otherwise they would have heard a bellyful from me about it. I'd castigate *them* here too, but that would be off topic. > Two: If randomization within the home is the issue, I'm not sure the > difference between a /48 and a /56 is all that significant. The point I wished to make is that we don't have a reasonable expectation that the size of a HOMENET subnet identifier will ever be constant over time and across renumbering events, much less across transfers between providers. We're not even confident that a HOMENET will even be offered *any* space for a subnet identifier. As a result, it means that Automatic Prefix Management here is basically unable to do it statelessly, i.e. by randomly generating subnet numbers from an identifier space of conventional size and testing for collision before using them. Any HOMENET autoprefix system will depend on the proper configuration of a central subnet identifier registry integrated or tightly coupled with the border gateway. When new link routers arrive, they have to ask the central registry for the next available subnet identifier and take out a lease on it, just like an IPv4 host that has to maintain a lease to use its interface addresses with DHCP. When link routers leave, they have to release their lease or the network must wait for it to timeout. And what happens when a HOMENET has four link routers in it, and the ISP renumbers the delegated prefix from a /60 to a /63? Obviously, something in the network has to kick two of the routers out of the HOMENET, but which two? Does the subscriber even get to choose? Basically, we've given up on stateless router autoconfiguration in HOMENET, and we're forced into a stateful solution. There are no good choices here, and the worst case outcome is that we will force the widespread adoption of NAT66 at HOMNET borders, precisely because it may turn out that subscribers want stable subnet identifiers, and more of them than their service providers are willing to provide at reasonable price. This all happened before, and we're not showing any signs of making sure it doesn't happen again. -- james woodyatt <j...@apple.com> core os networking _______________________________________________ homenet mailing list homenet@ietf.org https://www.ietf.org/mailman/listinfo/homenet