[ppml] The myth of IPv6-IPv4 interoperation, was: Re: Legacy /24s

[michael.dillon at bt.com]

> > Given that IPv4-IPv6 proxying is as simple to implement as 
> PNAT, one 
> > wonders why people are making such a fuss.
> 
> I think you mean IPv6 to IPv4 proxying? Letting IPv4-only 
> hosts talk to IPv6 is an impossibility. Unless CERN finish 
> the LHC and find the "answers to the universe" soon or 
> something that is ;-)

Proxying is always possible. Sometimes it mixes layers, but it is always
possible.

Let's give a concrete example. Joe's ISP in Lower Podunk has 80% of the
households on his dial-up IPv4 service. There is no broadband in Lower
Podunk and likely never will be due to the limited population, lack of
growth in the county, and generally rural spread of households. So Joe's
ISP has no plans to switch to IPv6 for at least the next 5 - 10 years.

Google acquires Facebook, launches the new GoogleBook service on the
IPv6 Internet only, and the state department of education announces that
all universities will be closed and higher education will only be
available on GoogleBook. Does Joe have a problem?

No, of course not. He puts A records for GoogleBook into his DNS servers
pointing at his V6 proxy. The proxy is actually hosted in the state
capital where IPv6 service is readly available. Joe's customers can now
use GoogleBook's IPv6 services via Joe's proxy server. Since GoogleBook
is implemented using the standard HTTP protocol, it doesn't much matter
whether there is IPv6 or v4 underneath. In fact, Joe's proxy functions
much like an IPv4 NAT box except that it uses unique IPv6 addresses for
each customer session rather than an address/port pair as in IPv4 NAT.

Operationally, you can expect people to just fix problems and make it
work, not agonize over whether or not the IETF could have documented
that solution 10 years ago.

--Michael Dillon