Eric Rescorla writes:
P.S. And btw, I'm not advocating NAT. What I'm advocating is that
we stop behaving as if we think that anyone who uses NAT is obviously
an idiot.
I don't think that I've seen anybody say that.
Most people who use NAT have no clue one way or
the other about NAT any more than they possess
clue about, oh say, 8086 instruction sets on
modern day processors. They aren't "idiots"
because they aren't claiming any (deep) knowledge
about the subject. They're merely ignorant and
just trying to get by with limited knowledge.
However, it's quite a leap to go from general
ignorance about the inner workings of computers
and networks to divining user's True Intents based
on their purchasing habits. I doubt that many
people go out with the intent of getting cancer
when they buy a pack of ciggies, but that is often
the end result. And people today have far more
clue about smoking than they do about NAT's.
People use NAT's for all kinds of perfectly
good reasons (I agree with Noel's analysis too,
btw), but it doesn't mean that cancer isn't an end
result there too. Voice is a great example of the
malignant chickens coming home to roost.
So I just don't see what the value is in
introducing the Voice of the Masses as indicative
of anything one way or the other about NAT. They
have requirements, and NAT's provide solutions for
some of those requirements, in much the same way
that smoking provides teenagers with a solution to
the requirement to be cool. Our job, however, is
to look at the *all* of the requirements and see
if they are best served with a given set of
technology. That's because we're paid to
understand this stuff and look out for the long
term health of the net. Which isn't to say that
we're infallible or incapable of indulging in
religious fervor, but we're at least semi informed
unlike the great unwashed masses.
Thus, I think it's *far* more productive to
actually limit the arguments to not whether NAT is
good or bad, but what requirements are we not
fulfilling which is leading people to NAT's as an
acceptible alternative (cf Noel's post). Some
people have obviously throw in the towel and would
like to have NAT's/ALG's to replace or augment the
end to end principle. Others aren't ready to give
up on end to end. I think it's premature to
declare the latter dead though. Some of the
requirements have simply not been hashed out to
any great degree. Fred's draft on operational
renumbering is a great example of something for
which there is more handwaving on our part than
actual experience. That doesn't mean that it's a
dead end though. All it means that it's something
that still needs to be dealt with. After all of
that is done can we actually have an intelligent
argument about what the right path is. This is,
IMO, quite literally a race.
Mike