[ppml] NANOG IPv4 Exhaustion BoF

[Joe Maimon]

David Conrad wrote:


> I would agree this scenario is far preferable to the first.  What  
> actually comes about is, of course, hard to determine at this point in  
> time.  I personally believe that policies should be explicitly  
> oriented towards promotion of the second view as one outcome of  
> transfers could be a market encouraging increased address space  
> utilization efficiency via NATs, which could reduce the pressure for  
> IPv6 deployment.


This would actually be a best case scenario, where access to ipv4 
internet is through ipv4 nat, but ipv6 is end-to-end.

Thus ipv6 service (which is plentiful and cheap) now has a clear 
advantage to it, one that will only grow as the rfc1918-ing of 
"consumer" customers picks up speed, and dual stack can actually work as 
a practical migration strategy.

This is actually the path of least resistance for ISP's wishing to 
migrate their customer base to ipv6 in the face of ipv4 shortage, 
whether or not ipv4 is available for whatever price.

Switch your users to pop boundaried rfc1918 natted to give "normal" 
access to ipv4 net, offer them all ipv6, offer the holdouts and 
content/service providers inbound ipv4 either sold by the port or by the 
address for a premium.

Market pressures and network effect should take care of the rest.

Without this approach, the pressure becomes all about enabling access to 
ipv4 from ipv6 (and ipv4 to ipv6), because otherwise ipv6 is useless.

So which is the better technology and outcome, dual stack of nat ipv4 
and end to end ipv6 or single stack ipv6 with {proxy | nat | undefined} 
solutions for ipv4 interop?

The situation for content providers is likely a bit more problematic 
than for ISP's.