Re: charter

[smd@ebone.net (Sean Doran)]

| I view it only as precluding approaches that are applicable only to
| IPv4. Are there any? 

ARP and other sticky protocols in the WAN (random routing with memory)
where the host rather than an aggregation of hosts is the unit of memory
or some other bottom-up scheme will probably work much less well in IPv6.  :-)

Before you laugh at that, consider RPF checks in multicast forwarding.

Hey, incidentally, did we restrict ourself to unicast IDR, or do we
have to solve the (still not fully solved for v4) multicast IDR too?

Oh see what a rathole this is?  Hence, let's solve the multihoming
problem for v6 with existing off-the-shelf technology, which is
nearly identical to v4.  Do that first.   

Next, try to figure out if we can come to some broad, rough consensus
on the technical design that shall come out of Ran's proposed design team,
and document that.   His goal is at least in part to 
propose an  agreeable split between "what" and "where", such that
hosts themselves don't care "where" they are as long as "what"
they're talking to doesn't change, and perhaps also to do a first
pass at a multihoming scheme which can take advantage of that split.

Now, should we constrain Ran to having his scheme coexist in some
particular way with CIDR-for-v6?   Probably we shouldn't formalize
that, although I do like the idea of a bit or two in the most significant
part of the unicast address to be used to indicate whether CIDR-for-v6
routers should IGNORE the prefix (i.e., neither use nor propagate).
This will facilitate experimentation by flipping one bit on whatever
comes out of the DNS on the send side, and recognizing that on the
receive side, and knowing that the first-cut CIDR-for-v6 system 
did NOT get involved in the building of forwarding tables in intermediate
systems.

| Note also, that the current charter doesn't say this WG will SOLVE the
| problem even for IPv6. It says we will start LOOKING AT approaches,

Now THAT's pragmatic!

| but we will then need a recheck (e.g., recharter and IESG approval)
| before we actually go work on a specific solution.

Well, I think we should actually get CIDR-for-v6 documented.
Randy had a whole mess of (private) questions before we started
using the list, that maybe should be answered. I noted some of
them in my summary-first-posting.

	Sean.