Re: Host-based may be the way to go, but network controls are neccessary

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

    > From: Iljitsch van Beijnum <iljitsch@muada.com>

    > It's unbelievable how much work is still going on on IPv6 which by all
    > accounts should have been deployed by now. ...
    > But IPv6 is still just IPv4 with bigger addresses.

Those who have known me in the IETF over the last 10 years will no doubt be
already aware that my reaction to these comments is trying to decide whether
to emit a scream of rage, or a scream of pain! :-)

    > So what's the right thing?

Alas, there is no answer to this question in its most global sense.

I can, relatively easily, tell you how to come up with a good technical
design; e.g. take a look at:

  http://www.isi.edu/newarch/

for an effort that's doing something like this, albeit at a higher layer. The
routing stuff has been done too (although some low-level parts, such as the
packet format, would benefit greatly from being updated to the age of
photons).

What I have no idea how to do is i) get the IETF to sign onto it (been there,
done that), or, and perhaps even harder, ii) get it deployed. The vast
majority of people just seem to love incremental tweaks, and only rarely do
you get the chance to perpetrate a revolution.


    > The original TCP/IP architecture assumes that an interface on one host
    > communicates with an interface on another host, the network always
    > knows what connects where and nobody will try to disrupt all of this.
    > Today, most services run on several hosts (load balancers) or the other
    > way around (NAT). Most of the network has no idea if destinations are
    > even reachable, let alone what the shortest path is (CIDR). Every
    > aspect of the network is open to constant disruption (DDoS et al.).

Yup.

    > Routing and layer 4 and up don't do what we need them to do so an
    > architectural overhaul is certainly in order.

Agree completely. (I could be snotty and say "and the IPv6 effort should have
done this", but that wouldn't be of any use to anyone.)

Sigh, I wish I had an answer to the non-technical part of the problem for
you. But I don't. I think I'll go work on my Japanese prints for a while.

	Noel