Re: IPv6 only host NAT64 requirements?

[Jen Linkova <[e-mail]>]

On Tue, Nov 21, 2017 at 10:44 AM, Brian E Carpenter
<[e-mail]> wrote:
> Fair enough. But now, and for many years, IPv6-only hosts** will
> be unusual (except on some cellular networks); IPv4-only hosts are
> common, and dual stack hosts are common.
>
> ** In my book, a host that contains even one IPv4-only application
> cannot be considered as an "IPv6-only" host.
>
>> People from the real world is coming to IETF saying 'we are deploying
>> Ipv6-only hosts' [1] [2]
>
> Sure. And I think the tests last week showed two things:
> 1. NAT64/DNS64 is viable for IPv6-only hosts (setting aside DNSSEC).
> 2. IPv6-only hosts are far from universal, even among IETF
> participants.

What I really wanted to know was exactly how universal those hosts are..

Unfortunately only 33 people have responded to the survey I sent ;( 8
of them said that they can not get their work
done on NAT64 while for the rest it was either "everything worked just
fine" or "smth was broken but not critical" (and yes, there was that
network issue which
affected the user experience..;(

It also showed that some applications which used to be IPv4-only are
now working on NAT64 network.
At least when you run the latest version of the app/the latest version of OSes.

I see it that way: I'm running a network and I do not want to run a
dual-stack one. At least I'd like to minimize IPv4-only/dual-stack
surface.
The fact that there will be some old hosts/apps which could not work
on v4-only network just means that I might not be able
to migrate *all* hosts to v6-only overnight/any time soon.
The same way I still keep 2.4Ghz SSIDs (and some devices using it) in
addition to 5Ghz ones (where majority of devices are sitting).
Number of Ipv4-only applications will go down overtime and eventually
it would not have much sense to maintain them
(as it would be cheaper to get rid of them than maintain dual-stack
only for a few old things..).
When that moment comes for a given network depends on many factors indeed.

> And BTW thanks for the effort behind ietf-nat64. I think it
> was a very valuable effort.

You are welcome, I'm glad to hear that people find this experiment useful.
All credits goes to the IETF NOC actually, I was just a messenger ;)
(and someone who enjoyed running Wireshark ;))

-- 
SY, Jen Linkova aka Furry