Re: [v6ops] IPv6 mostly for DS-Lite

[Ole Troan <[e-mail]>]

>> The market has picked both DS-lite and DNS64/NAT64. Both are deployed by ISPs.  One actually worked. The other required a series of corrections to be put out making it more visible to the IETF.
> 
> I don't think anyone is saying "switch them off". It's a question of where we spend IETF effort.
> 
>> Daisy chaining DS-lite and 464XLAT will be a real mess as both use the same IPv4 address rage for source addresses.  
> 
> That is... unfortunate.

In the example where the hosts gets IPv4 via 464XLAT and the site gets IPv4 access via DS-lite, then it should be better if the host used the ISP NAT64 instance instead of the CPE NAT64 instance.


> 
>> Or are you saying only those that can get full dual stack can deploy IPv6-mostly.
> 
> I'm not sure I can parse that. The end point must be that applications *see* a full dual stack whatever mechanism is deployed on the wire. That applies to every host on an IPv6-mostly network (unless there is such a thing as an IPv4-only host, I suppose).

464XLAT in itself has 2-3 NATs. Chain it with DS-lite, you would have another 2? Nice! :-)
_If_ all hosts (very unlikely of course) supported 464XLAT then the CPE could of course be IPv6 only.
Alternatively, the CPE could act as a CLAT and proxy for those IPv4 only nodes that don’t.
At least that’s a path to a single mechanism.

Cheers,
Ole