Thomas,
I think you're actually both right.
Noel's comment was meant as a general case for even a very large
enterprise. You provided a counterexample of some number of providers that
will need huge amounts of address space. Much as I dislike NAT I believe
it's an unproven assertion that those providers won't find a way to molest
the architecture further to make it work. Really all that is required is
some small percentage of public addresses, based on usage patterns. Any
Bell Heads want to take a stab at that percentage?
My hope is, however, that a NAT deployment would be 6-4, so that they could
at least have the pipe dream that those NATs might some day become
unnecessary.
Also, many people have made claims of grand visions when it comes to usage
of address space, and these grand visions have been rebutted with "it
hasn't happened yet." In some cases it hasn't happened because it couldn't
happen. Fred tells the story of China. I can tell the story of a cable
provider, back in 1993, who were planning a large scale Cable-IP
deployment. They were ahead of their time, but not by that much.
Eliot