Re: draft-ietf-multi6-multihoming-requirements-06.txt

["J. Noel Chiappa" <jnc@ginger.lcs.mit.edu>]

    > From: Jay Ford <jay-ford@uiowa.edu> 

    > The problem I see is that (srcaddr,dstaddr) selection by a multihomed
    > IPv6 host in the context of provider-based addressing can preempt the
    > routing. That means that the loading will be determined by clueless
    > edge hosts, which is about the worst place for such decisions to be made.

First, I'm reminded of Churchill's famous line about democracy - about how it
was the worst form of government, except for all the others. Similarly,
having the edge-hosts make routing decisions seems like the worst place to do
it - except for all the other choices!

I don't understand why people get so worried about having the hosts be more
involved in routing - but then perhaps my perspective helps me. E.g. way back
when (i.e. the ARPANet), the network did congestion-avoidance, not the hosts,
in part because "we have to protect the network and oh of course it's too
complex for hosts to do". But having the network do it for the hosts turned
out to have disadvantages. Now every host has very fancy congestion-avoidance
algorithms in it, and it works fine, and everybody thinks its normal. Also,
every human being (well, almost :-) is perfectly OK with the concept of
hopping in their car/whatever and reading a map to get somewhere. It's not
quantum mechanics, people.


Second, you are right that (srcaddr,dstaddr) selection is something of a
routing decision. However, making that choice more intelligently requires a
far better routing architecture than the one we have now - one that is
prepared to provide more information to hosts to allow them to make those
choices. (See Figure 1.) Lacking that, flipping a virtual coin (perhaps a
loaded one) is about as good as we're going to do...

	Noel