On Oct 7, 2010, at 10:20 AM, Noel Chiappa wrote:
From: Brian E Carpenter <brian(_dot_)e(_dot_)carpenter(_at_)gmail(_dot_)com>
The problem is that the creation of disjoint addressing realms (due to
NAT and to IPv4/IPv6 coexistence) has made distributed application
design almost impossible without kludges.
See, this is the kind of thing I was talking about in my early post in the
recent incarnation of this thread. Complaining about the existence of
disjoint naming realms, and how it has complicated our lives, is like a
rocket scientist complaining about gravity, and how hard it makes their job.
(OK, it's not quite a perfect analogy, since gravity is fundamental, but I
think my point is clear.)
They were inevitable, end of story.
No, they were not inevitable, at least not in the long term. Two cases:
1. There were IPng proposals that would have made IPng an extension of IPv4
space, and allowed (at least in the interim) all IPng traffic to be tunneled
over IPv4 networks. This understandably made routing people and network
operators uneasy as nobody wanted to live with the Class C swamp forever. But
in hindsight there were probably other ways of dealing with that problem. And
being able to leverage the existing IPv4 network in a general way would have
made IPng easier to deploy. (Admittedly, what looked deployable in the early
1990s when these tradeoffs were being discussed is very different from what
looks deployable now. In 1991, say, it was much easier to imagine the whole
Internet migrating to a new protocol than it was even a couple of years later.)
2. In the late 1990s when it became apparent that NATs were causing problems,
IETF had an opportunity to put a stake in the ground. It could have tried to
find a way to integrate NATs into an explicit Internet architecture; it could
have explained why NATs presented difficult problems for which there were no
good solutions; it could have tried to define NAT in such a way as to permit a
more graceful migration to IPv6. It failed at all three of these, not because
of insurmountable technical difficulties, but because (a) too many people were
afraid of alienating NAT vendors, and (b) IAB, the only part of IETF that had
any responsibility for architectural direction, had been stripped of its power
in the wake of Kobe.
In fact, as we should realize by now, IAB's concerns that dictated the Kobe
decision were extremely well founded, even if a lot of us (myself included)
didn't like the specific choice they made. But as it turned out, we didn't
really have enough time to use our normal "rough consensus" process to specify
IPng.
All of this is water that has already flowed under the bridge, of course. But
I think there is at least one thing that we need to learn from this going
forward, and that is that there really is a need for a small group of wise and
widely-trusted people to be able to set architectural direction for the
internet. And occasionally, they're going to make unpopular decisions. But
those are the sort of issues for which a rough consensus process demonstrably
does not work.
Keith
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf