Re: [rfc2462bis issue 274] conflict between MLD and NS delay

[JINMEI Tatuya / $B?@L at C#:H(B <jinmei at isl.rdc.toshiba.co.jp>]

>>>>> On Tue, 10 Feb 2004 20:31:19 +0900 (JST), 
>>>>> itojun@itojun.org (Jun-ichiro itojun Hagino) said:

>> 3. What happens if someone builds a stack from multiple vendors
>> (e.g. MLD from vendor X and NDP from vendor Y)?

> 	this is "someone"'s responsibility to ensure the documented behavior,
> 	IMHO.  it is not a concern for standard document (no need to document).

I basically agree.  IMO, this kind of thing is implementation
dependent, and can be out of scope of a specification document.

Regarding our own implementation (KAME/BSD), it joins the
solicited-node multicast group corresponding to a link-local unicast
address within the kernel; it doesn't need a separate system call
(IPV6_JOIN_GROUP) to join the group in this particular case.  So, I
believe it's not so difficult to implement the procedure I showed in
my previous message at least on one OS that I know of.

We could add a note in rfc2462bis that in some OSes it may not be
effective to separate the join process and the listen process, though
I personally think it's overkilling in this document.

					JINMEI, Tatuya
					Communication Platform Lab.
					Corporate R&D Center, Toshiba Corp.
					jinmei@isl.rdc.toshiba.co.jp

--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------