Re: [DNSOP] [[e-mail]: I-D Action: draft-grothoff-iesg-special-use-p2p-names-00.txt]

[David Conrad <[e-mail]>]

On Dec 3, 2013, at 9:27 AM, Ted Lemon <[e-mail]> wrote:
> On Dec 3, 2013, at 12:08 PM, Stephane Bortzmeyer <[e-mail]> wrote:
>> If we want actual testing of the ability to run non-IN classes, I
>> accept donations in bitcoins to do so in my lab :-) But, anyway, you
>> have very little chance of convincing any developer to spend time in
>> this direction, which is clearly dead.
> 
> Furthermore, it doesn't even address the problem, because the problem we are talking about is a naming problem, not a problem that exists within the DNS protocol.  So by definition you can't solve it with classes, because classes are a DNS thing.

If it isn't a DNS thing, then why is there a discussion of the allocation of top-level _domains_?  If these applications aren't making use of the DNS, then why does anyone care if they use something that looks like a domain name?

The issue here is that they are, in fact, using the DNS in the sense that they are using applications that expect to query a local DNS stub resolver and they are (presumably) expecting that resolver to distinguish between a query for a "regular" domain name and one of these non-DNS "domain names". The reason they can't use classes is because there isn't a convenient mechanism to specify an alternative class using the vast majority of applications, hence a fallback to distinguishing using "special" names.

I have significant concerns that this is not going to scale. There needs to be a better discriminator than simply "because I wrote an RFC that describes a protocol that assumes a top-level domain".

Regards,
-drc