[arin-ppml] v4 to v6 obstacles

[michael.dillon at bt.com]

> > Agreed. That is why I put a ? on the 80%. If some really big and 
> > important players (applications) go dual stack, that covers 
> a lot of 
> > territory.
> 
> You mean like:
> 
> Google
> Yahoo
> MSN
> 
> (All of whom have publicly announced dual stack plans)?

This is meaningless.

It is one thing to "go dual stack" on your core network
architecture that rarely grows in number of devices, and
which has enough spare IPv4 addresses in existing blocks
to last another 5 years. But it is an entirely different 
thing to "go dual stack" on an edge architecture that is
continuosly growing and which need an infusion of new 
IPv4 addresses every 6 months to handle growth.

I think that it is good for providers to dual stack their
cores, but I think that the decision to dual stack the edge
is one that should not be made lightly. For a provider
like Yahoo, Google and MSN, where the edge of the network
is the core of their business (i.e. the insides of the
data center are the edge of the network), it certainly
makes sense to dual-stack their core network while relying
on NAT in the data center (or IPv6 with NAT-PT and 6to4)
for their core business. However this is a very different
topology and business scenario to the one that network
operators are grappling with.

--Michael Dillon