From: Keith Moore [mailto:moore(_at_)cs(_dot_)utk(_dot_)edu]
there is one important class of bad ideas that doesn't go
away in IETF -- the class of bad ideas that is obviously bad
from a wider perspective but which looks good to a set of
people who are focused on a narrow problem. and in IETF what
we often do with those ideas is to protect them and encourage
development of them in isolation by giving them a working
group. we sometimes even write those groups' charters in
such a way as to discourage clue donation or discussion of
other ways of solving the problem.
That is a somewhat cynical way to describe IPSEC isn't it? Care to mention
any other groups that fit that description?
DKIM comes to mind, as does zeroconf. But I've seen so many examples
of this over the years (including IPsec) that I've lost track.
The IESG and the IETF in general has hardly demonstrated an infalible
understanding of what is and is not a bad idea, nor for that matter has
true. but the fact that we're not infallable doesn't mean we shouldn't
try to improve things.
No Keith, you are not Vint Cerf, or Tim Berners-Lee
and neither are the real Vint or Tim, respectively. (both smart guys
whom I respect, but there's a difference between any real person and his
reputation. and this isn't an discussion about personalities, it's a
discussion about how to do protocol enginering)
I know that folk focused on narrow problems have tended to come up with
narrow solutions. That is hardly suprising, the rules of engagement here
prohibit the discussion of the general.
which is my point - we need to change the rules of engagement.
Take DKIM for example we are about to discuss a one off policy language to
serve a single protocol, not because there is only a single protocol that
requires policy but because there are people in the establishment who tried
policy fifteen years ago, failled to solve the problem and have declared it
'insoluble'. There is also the problem of the other group who need s to be
part of the policy discussion which has repeatedly demonstrated itself to be
unwilling to listen to any outside view. Try to explain a problem to them and
its 'la la la I'm not listening'.
DKIM as currently described in the I-Ds is a lot more broken than that,
but they're not listening either. But it's a lot bigger problem than
any single working group. I find DKIM a convenient example in this
discussion because it's current, and because of my long history of
working with email I have a keener interest in that WG than most.
But it's not hard to find other examples.
Ietf mailing list