(Forgive me for the late comment, too) >>>>> On Sat, 26 Jun 2004 21:30:19 -0400, >>>>> "Soliman Hesham" <[EMAIL PROTECTED]> said:
>> as the Subject; line of the email says, once you define >> host/router >> be per-interface thing, you will need to describe >> router behavior for >> every possible combination of mixture of interface. i >> still think >> we did not finish describing simple all-host-interface >> behavior nor >> all-router-interface-without-virtual-router behavior, i strongly >> object to the change. > => ok I don't see that need to describe forwarding between > interfaces in 2461bis. I also prefer to make a clear definition > than leave it empty. I'll let others say what they think > before I keep or remove the text. > I don't think there is a doc that describes a host/router > for v4 or v6. I find it difficult to see the reason for such > doc. I tend to agree with itojun in that it is not so trivial to define the "mixed host/router behavior". We have once discussed related points in the context of rfc2462bis. See, for example, the following messages: http://www1.ietf.org/mail-archive/web/ipv6/current/msg01521.html http://www1.ietf.org/mail-archive/web/ipv6/current/msg01524.html http://www1.ietf.org/mail-archive/web/ipv6/current/msg01525.html As pointed out in msg01521.html, the notion of "the per-interface router" has not been well defined. RFC2460 talks about something on this, but it's still not clear whether a "mixed router" can *forward-to* a host-side interface (see msg01525.html). Perhaps the clear definition itself is not in the scope of rfc2461bis, but I don't think it makes sense to discuss how the mixed behavior is configured in rfc2461bis (e.g., with the "IsRouter" variable Hesham proposed) before clarifying the definition. Regarding rfc2461bis, if we allow the mixed behavior, we also need to clarify (at least) some other things that include: - whether a mixed node can receive an RA on a "host" interface to configure a default router list - if it can, how the configured default route interacts (or does not interact) with other routes used for forwarding on "router" interfaces. - whether a mixed node can receive an ND redirect on a "host" interface to adjust the route to a particular destination - if it can, how the adjusted route interacts (or does not interact) with other routes used for forwarding on "router" interfaces. Considering the lack of the base definition of the "mixed behavior" and the complexity of the additional clarifications, I'd rather concentrate on "all host" or "all router" nodes in rfc2461bis. And for that matter, I reached the same conclusion in rfc2462bis and edited the 02 version that way (after confirming the decision on the list). Meanwhile, if we concentrate on the "simple" cases, I think we need to clarify one thing. RFC2461 has the notion of "advertising interfaces". An advertising interface is to control whether to send router advertisement, and is, of course, defined per-interface basis. It seems to me the existence of this notion is one source of the confusion on whether the ND specification allows the mixed behavior or not. So, (if we concentrate on the "simple" cases), I think we should emphasize that even if an interface is not an advertising interface the node still acts as a router on that interface (e.g., it can forward from/to that interface, exchange routing information on that interface, etc). One last point: I hear that some implementations explicitly support the "mixed behavior" (perhaps the Windows implementation?). We may (or may not) need to care about those existing implementations to edit wording for rfc2461bis (at least we will not want to "invalidate" those by this revise work). JINMEI, Tatuya Communication Platform Lab. Corporate R&D Center, Toshiba Corp. [EMAIL PROTECTED] -------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
