IPv6 PI allocation

[Doug Barton]

Nick Hilliard wrote:

> If we stop PI6 assignments, then we would have a situation where we
> had a functional ipv4 multi-homing solution, but no functional
> ipv6 multihoming solution.  This in itself would kill ipv6
> deployment, because ipv4 would then have a significant technical
> advantage over ipv6 for end-users.  Why regress to something less
> functional?

In addition to the technical arguments which I think Nick has argued
very eloquently, there are also serious economic arguments that just
can't be ignored. PI space in IPv4 is something that large scale
Internet businesses (and some smaller ones) _depend_ on (whether their
reasoning is misguided or not). I have talked to several who refuse to
even devote R&D time to IPv6 until there is PI space available. They
simply cannot make a case for a technology that "regresses" in this
area. I realize these arguments have been had before, I've been at
many of the RIR meetings where they were discussed. They go along very
predictable lines, many of which have been repeated here. But at the
end of the day, Nick (and others in this thread) are right. If you
want to get IPv6 off the ground, you have to give the people what they
want.

Colm MacCarthaigh wrote:

> Exactly, I actually think the best thing we could do to mitigate 
> this problem would be to mandate that the smallest network size 
> MUST be /64, that way routing vendors can concentrate on designing 
> fibs that hash-bucket at that boundary.

BZZZZZZzzzzzzzzzzzzzzzzztttttttttttt! See above on the "give the
people what they want" bit. They tried this with the "smallest network
MUST be /48" and that didn't fly either. No matter how smart you think
you are, and no matter how good your reasoning is to do what you think
is right, there is always going to be someone who has a better (or at
least different) idea, and at the end of the day, they are going to
use the TECHNOLOGY in the way that suits them anyway. So trying to
mandate your good ideas just isn't going to work.

> Right now, the use of /127's and /126's for link address by some 
> networks are a bigger part of future problems than anything 
> represented by PI.

I actually find this paragraph really interesting in light of what you
say below. If there is a problem, it's a technology problem, so it has
technology solutions.

> Ultimately there are no magic engineering limits that prevents 
> anyone from routing or switching billions of routes, it will 
> ultimately cost us in terms of more sillicon and memory, but we 
> only need to do a better job than IPv4 managed, it's no big deal.

This I agree with completely. We not only need to learn from the
mistakes made in IPv4 (*cough* classful routing *cough*) we need to
learn from what worked.

Doug

-- 
     If you're never wrong, you're not trying hard enough