On Fri, Oct 7, 2011 at 14:25, Erik Nordmark <nordm...@cisco.com> wrote:
> The slides say: > "Support arbitrary topologies including loops" > > What is the implication of that on IPv4? > Are you assuming that the dual-stack home network will delegate IPv4 > prefixes and route IPv4, with no internal NATs? > > How would someone transition to such an IPv4 network from their current > daisy-chain or tree of IPv4 NATs? > IPv4 won't work. You would likely never get to such a topology (or never stay there) if you need IPv4, since you would notice it's not working and unplug it. :-) However, it would be nice if IPv6 were able to work in this situation as well, so if you don't need IPv6, you can plug things in any random way and have things work. Supporting arbitrary topologies and loops is not an absolute requirement (especially loops, since loops will break IPv4), but it would be nice to work in multiple topologies, and you have a routing protocol anyway then being able to support loops comes for free. The other question is about the assumptions about the requirements of > stability of the prefixes that are assigned to links. > > For routing you have: > - Survive loss of any router for any time period > - Works regardless of router boot order > > That combined with your strawman prefix assignment seems to result in the > set of prefixes assigned to a link changing each time a router goes down. Am > I reading things correctly? > To achieve stability, routers would need to write the prefixes they form into stable storage (like they would a DHCP lease). I think it's reasonable to write to stable storage whenever you configure a new prefix or change an existing prefix (but not if you simply refresh it). If the router comes back up and nobody has the prefix, it would simply claim it back. If someone else has it, it would have to just form a new one using loop detection. Cheers, Lorenzo
_______________________________________________ homenet mailing list homenet@ietf.org https://www.ietf.org/mailman/listinfo/homenet