DNAME issues (was Re: ip6.int deprecation)

[Iljitsch van Beijnum]

On 9-mei-2006, at 11:59, Geoff Huston wrote:

>> Wouldn't having a DNAME record for ip6.int be a very good way to
>> uncover DNAME brokeness without real trouble? After all, people who
>> still do ip6.int aren't going to get what they're looking for in the
>> alternative anyway.

> If you want to look at IPv6 as a continuing technology testbed for  
> the "real" deployment elsewhere or elsewhen,  then there is much to  
> be said for this type of approach of using IPv6 as a test rig for  
> novel approaches to provisioning of infrastructure services.

Huh?

Since when does DNAME need testing? It has been out for years, and if  
the IETF had still been in the business of advancing standards, it  
would have been one by now.

There are basically two sane ways to operate: do the right thing, or  
do the thing that works best. Since in this case the fallout of _not_  
doing the thing that works best (= work around broken  
implementations) is negligible, why not do the right thing? (= use  
the tools that we have as effectively as possible)

The way things are going now we can never deploy DNAME because there  
are broken implementations out there and people don't know/care  
because we can never deploy DNAME. That kind of stuff annoys me to no  
end.

> On the other hand, or course, there is a significant body of  
> opinion that wants to get over the concept of IPv6 is still an  
> experimentation rig and promote the perspective that Ipv6 is  
> stable, functionally mature and ready for deployment, and further  
> experimentation and testing in infrastructure elements is not  
> precisely the current intended role for IPv6.

That's nice and I'm not actively disagreeing here, but as long as I  
can't autoconfigure a DNS resolver address on my system on pretty  
much any of the IPv6 implementations out there this significant body  
may want to steer clear of horses quite this high.