Re: draft-bourbaki-6man-classless-ipv6-00

[james woodyatt <[e-mail]>]

On Jun 2, 2017, at 10:14, Manfredi, Albert E <[e-mail]> wrote:
> On Jun 2, 2017, at 09:08, Lorenzo Colitti <[e-mail]> wrote:
>> 
>> Example: I don't want us to have to deal with NAT any more, ever.
> 
> I've seen this comment made before, and yet the /64 hard boundary practically guarantees that NAT will get used with IPv6. It's exactly the same solution, for exactly the same problem. Inability to expand at the edges.

I have reluctantly come to believe that service providers MUST have the ability to force their economy class subscribers into artificially small allocations, so they can charge premium prices for larger ones drawn from the same functionally limitless supply at equivalent cost. We have to expect that only business and luxury class subscribers will have sufficiently large allocations that IPv6/NAT will add more costs than it saves them.

I once believed IPv4/NAT was about preserving scarce number resources. I now think that was wrong. It was always about optimizing rents across differing classes of subscribers even when the extracted resource on which the rents are derived is practically unlimited. In this one respect, IPv6 is no different than IPv4. As a purely technical matter, the rent MUST flow.

Accordingly, I say we should stop arguing that IPv6 does not require NAT. For certain classes of subscriber, NAT is essential for making IPv6 service available at an affordable price. Indeed, any internet layer that uses fixed size global scope addresses requires NAT at the provider/subscriber boundary.


--james woodyatt <[e-mail] <mailto:[e-mail]>>