Anybody behind a NAT64?

[Tore Anderson]

* Daniel Roesen

>> That said, I'm expecting that due to IPv4 exhaustion, most of the 
>> providers will eventually have to switch to private IPv4 addresses
>> + CGN. When that happens, avoiding DNS64 doesn't really matter all
>> that much anymore, since all the IPv4-destined traffic will have 
>> to pass through a NAT instance either way.
> 
> Well, it avoids the inherent problem of translation between 
> incompatible addressing systems - see all the NAT64 failure cases.

As I understand it, the NAT64 failure cases stem primarily from
applications not using DNS[64] to begin with, or that use IPv4-only sockets.

At least in the mobile world. Cameron please correct me if I'm wrong,
but with T-Mobile's 464XLAT service, DNS64 is always used. So
applications that use DNS will prefer to bind to IPv6 sockets when
communicating with IPv4 destinations. In spite of that, as far as I
know, the architecture solves all the known application problems
experienced with IPv6-only+NAT64/DNS64,

> With "DNS64 desired" signalling, you could avoid NAT64 problems for 
> IPv4-capable clients while at the same time support IPv6-only
> clients via NAT64.

I can see that. Question is whether or not it's too late to matter.

-- 
Tore Anderson