> Additionally, adopting this definition also opens up the possibility > of a "half-host, half-router" node, like: > > ------(I1)Node(I2)------- > (I3) > | > | > > where I1 and I2 are "normal" interfaces for which the node is acting > as a router, and I3 is an interface to a "back-door" for which the > node is acting as a host. This node can accept a router advertisement > on interface I3 to configure an address or even configure a "default > route" to I3 (though the latter part is out of scope of 2462bis). > > Should we, if adopting the per-interface definition, explicitly > mention this type of configuration in rfc2462bis?
Is there a proposed definition for the per-interface router? I'm asking because we want to avoid having "forward from" and "forward to" interfaces - if the node is a router on a interface it should be able to forward packets both from and to that interface. And if a node is not a router on an interface it should neither forward packets (not explicitly addressed to itself) from that interface nor to that interface. If we decide to add the per-interface definition I'd suggest actually making that a separate document (which can be folded into future RFC 2460 revisions) instead of "hiding it" in some other document. I think your above concern about the default route on I3 can be reduced to a previously unspecified problem; the RFC 2461 silence on multihomed nodes. Per RFC 2461 the default router list is per interface thus there would not be any harm (in the conceptual model) to add that default route on a node that has other interfaces which are router interfaces. If one were to specify how RFC 2461 (and 62) apply to multihomed nodes then one would be faced with this issue. Of course, the above is a bit dishonest because 1) many implementations support multiple interface with 2461 even though the specification is silent about them and 2) I suspect many of those implementations have a global - and not per-interface - default router list. So do we or do we not want to 1. specify the per-interface router definition 2. specify how RFC 2461 (and 62) behave on a multihomed node Erik -------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
