Re: NOTE: WG Last Call: draft-ietf-v6ops-mech-v2-01.txt

[itojun@itojun.org (Jun-ichiro itojun Hagino)]

> Hmm - you seem to be assuming that the only reason for router A
> to send an RS to router B (and get an RA back) is for the purpose
> of autoconfiguration (i.e., RFC 2462, section 5.5.3). I'm not
> assuming that at all.
> 
> I believe it is useful in some cases for router A to discover the prefixes
> being advertised by router B w/o necessarily seeking to autoconfigure
> addresses based on those prefixes. (RFC 2461, section 6.2.7 gives one
> such example.)

	if
	- A advertises P1/64 and P2/64, and has P1:id1 and P2:id2 as address
	- B advertises P3/64, and has P3:id3 as address
	onto the same ethernet broadcast domain, A would think that P3::/64 
	as off-link, and sends packets to P3:id3 to default route
	(instead of performing NS/NA).  we can cook up more complex example
	like this, but i wouldn't bother.  we really need simplification here.

	i personally would like to see RFC2461 section 6.2.7 to be updated to
	require routers to advertise consistent prefix information.  we can
	consider it an operational issue (instead of implementation issue) so
	backward compatibility should not matter that much.

itojun