Re: Old / new NAT-PT coexistance?

[Iljitsch van Beijnum <iljitsch@muada.com>]

On 30 jan 2008, at 19:13, Iljitsch van Beijnum wrote:

> The new style that Brian and I are working on is going to require  
> changes on the IPv6 side, which won't be made over night: looking up  
> A records by the hosts themselves rather than having a DNS ALG  
> generate synthetic AAAA records for IPv4-only destinations.

> The problem is that you would have to provision different DNS  
> resolvers for hosts that are IPv6-only and do not support the new  
> style NAT-PT versus the ones that do and dual stack hosts. This  
> could be a problem. Solution: have hosts that do DNS lookups over  
> IPv6 transport indicate that they support new style NAT-PT and don't  
> need the synthetic AAAA records through EDNS0.

Another option would be for the server to simply always generate the  
synthetic AAAA records as per existing NAT-PT, but then add an EDNS0  
option that identifies the synthetic addresses. Hosts that are aware  
of this then know they're really talking to IPv4 destinations and  
could present ::ffff:x.x.x.x addresses (or x.x.x.x addresses) to  
applications so applications can adjust their behavior to what's  
needed for IPv4 operation.

This also nicely avoids the whole issue of how to configure/discover  
the /96 prefix used by the translator. The downside is that there will  
be significant old style NAT-PT and there is still the chance of  
leakage of the synthetic AAAA records as long as DNS servers don't  
implement the new EDNS0 option and filter out these records.

Thoughts?