> But this is just an operational procedure, which needs to be stated in
> the multihoming solution documents, but something that IMHO must not
> be added to the ICMPv6 spec, as it's really out of scope from ICMP
> perspective.

I agree with Pekka. The purpose of the ICMPv6 document is to reserve the
code and define the format. As Marcelo said, the MH code can work with
this code and format. Let's just go with it.

I understand Marcelo's concern. He (and I) would like to make life as
good as possible for hosts in multi-homed sites. This will require
operational rules for hosts and for routers. As I mentioned in a
previous message, the code definition is a nice first step. It tells the
host that it should try another source address. It is only a first step,
because it does not tell which one. The host will have to use some
heuristic. In some cases, e.g. when there are just two addresses, the
heuristic is trivial. In more complex cases, having more information
might help. But it may be a bit early to determine how this information
should be passed. We could pass it in ICMP messages, but we could just
as well use a yet-to-be-defined information protocol associating exit
routers with allowed prefixes, or even let hosts use some cached
knowledge from previous connections. Given the wide choices, I would
rather not try to mess around with the ICMPv6 specification.

-- Christian Huitema

--------------------------------------------------------------------
IETF IPv6 working group mailing list
[EMAIL PROTECTED]
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to