On Thursday, May 1, 2003, at 14:29 US/Pacific, Peter Deutsch wrote:
[...] As you pointed out earlier, folks talked about separating out
point identifier in the past, maybe we should reexamine this suggesion
in light of today's impasse? Maybe what we really need here is to
one of the axioms and see if it takes us to a non-Euclidian geometric
I am reminded of a pleasant discussion I had on Wednesday with one of
my colleagues, who studied under David Cheriton at Stanford. He
recommended Dr. Cheriton's TRIAD paper to me, which takes pretty much
exactly the approach you suggest above, so I dutifully scurried off and
TRIAD is a wonderful idea. I love it. I'm almost sold. Almost.
Unfortunately, while I see how the approach can be made to work without
a mass rollout of new functionality in the routers, or even in the
hosts, it *does* require a new infrastructure of relay agents. In
other words, his fix for NAT doesn't work, because the existing NAT
deployments interfere with it. Doh!
So I sat down with a hot cup of coffee in the atrium and sketched up an
idea of what it would take in the hosts (and in strategically
positioned middleboxes in the network) to work around the problem posed
by existing NAT deployments, so that we could phase them out and get
back the all-singing, all-dancing end-to-end goodness that is supposed
to have existed in the ancient Internet world.
I've read RFC 3424, and I'm pretty familiar with the technical issues
it describes, as well as the various solutions that have been tried.
Ultimately, I ended up reinventing Christian Huitema's Teredo proposal,
only-- the thing I was thinking about tunneling over UDP/IPv4 to solve
the RFC 3424 problems wasn't IPv6, it was a horrible abomination cooked
up to try to avoid deploying IPv6. Again: doh!
For a brief moment, I thought it might be helpful to make the following
suggestion: maybe the problem with Teredo is that the *only* thing you
can do with it is extend IPv6 service over IPv4 through reasonably
well-behaved NAT implementations. What if we were to specify a
"generic Teredo" that concentrated solely on solving the RFC 3424
problem in a generic way and let the market decide on a profile of the
"generic Teredo" service that it likes?
And then I sobered up and realized that an idea that seemed so
eminently sensible to me would likely be regarded by the IAB as only
slightly less welcome than a cholera outbreak.
Nah, let's just continue trying to square the circle....
A fine idea. I have my square right here.
j h woodyatt <jhw(_at_)wetware(_dot_)com>
that's my village calling... no doubt, they want their idiot back.