Re: (ngtrans) Joint DNSEXT & NGTRANS summary

[Johan Ihren <johani@autonomica.se>]

Jun-ichiro itojun Hagino <itojun@iijlab.net> writes:

> >Ok. Before I repeat the mistake of guessing at your configuration I
> >suggest you tell me about it.
> >
> >Personally I use a forwarding config for v6-only nameservers and I
> >believe that this is by far the most common case. 
> 
> 	I have been talking about a nameserver, with root.hint, on
> 	IPv4/v6 dual stack machine.  to be more precise, see below.
> 
> 	I never have talked about nameservers on IPv6-only machine,
> 	they will never work without some help from IPv4/v6 dual stack
> 	nameserver.  (and it is rather unrealistic to talk about
> 	nameeservers in IPv6-only network today...)

Ok, we were clearly talking about different things. And this further
clarifies the initial difference in opinion on the size of the
deployed base of full-service resolvers that are IPv6-aware and 
-caring.

The issue with roots available over v6 transport was mostly a side
track that was relevant only for the subset of full-service resolvers
that are v6-only.

As to whether it is unrealistic to talk about name servers on
IPv6-only networks or not, I still argue that although they are a
small base today (which we now agree upon) they will be the dominant
species at some point in the future. And now is the time we have to
decide how the world they will populate should work.

But the core issue is still whether it is too late to upgrade the
deployed full-service resolvers to do AAAA synthesis or not. And if
not too late then how long it will take.

I think that given a clear decision (which seems impossible) this
could be fairly rapid for the full-service resolvers that actually see
real traffic since they are almost certainly to a large degree waiting
for the outcome of this A6/AAAA commotion.

And given A6, that would not delay deployment [of IPv6], since new
deployment would have the benefit of doing it right [i.e. A6] from the
outset thereby rapidly creating even more incentive for the installed
base (i.e. the dual-stack or forwarding servers, all of them bind9) to
upgrade. 

It is important to weigh in the difficulties of pushing a deployed
base to upgrade. But in this particular case I worry that that problem
unnecessarily may be overshadowing the design issues.

Anyone that presently operate a dual-stack server or a forwarding
server with only v6 transport and isn't aware that there are unsolved
problems with DNS for IPv6 and that it will in all likelihood be
necessary to change configs/upgrade software/modify tools/etc is not
really on top of his or her responsibilities. 

Fortunately, since we are still in the early days, I really think they
constitute a small minority. The rest, regardless of their position on
the A6/AAAA issue, are at least fully aware that such issues *exist*
and hence aware that they will possibly have to change stuff somewhat
as those issues get resolved. The v4-only-dont-care-about-v6-although-
their-resolver-groks-aaaa is a different story, but I care less about
breaking lookups of v6 records for them.

So, to my mind, A6 would not necessarily delay deployment of IPv6
significantly. What is clearly delaying deployment is lack of consensus.

But I don't know. It has to be a judgement call.

Johan