Re: [Inip-discuss] Domain Names

[Suzanne Woolf <[e-mail]>]

On Jan 22, 2016, at 5:49 PM, Andrew Sullivan <[e-mail]> wrote:

> On Fri, Jan 22, 2016 at 08:40:28PM +0000, Edward Lewis wrote:
>> I should add one anecdotal story to this.  One of the "legendary figures
>> of DNS" once told me that a major screw up was placing the 16-bit field
>> for CLASS after the field for the "owner name".  Had the 16-bit field come
>> first, it could have been interesting, a way to distinguish different ways
>> to encode the owner name.  Being afterwards, one had to stick to one lone
>> means of encoding the owner name in order to discover the CLASS value.
>> I'll stress that I got this as part of an oral history, not documented, so
>> it's almost worth the paper it's written on.  But it is what makes me call
>> CLASS vestigial and a DNS artifact, not a Domain Name core concept.
> 
> The unwritten draft in my head, "CLASS Considered Useless", will make
> exactly this argument.  BTW, it's not just oral history.  I have found
> this argument in the very earliest part of the namedroppers archives.


Part of the problem here, it seems to me, is that we've gotten this far without a definition of "domain name" except what we can find implicitly in certain protocol definitions-- but nonetheless there's a concept that has developed over time, one attribute of which is that it's consistent in certain desirable ways across multiple possible resolution protocols. This assumption is being used in new protocol development, without being explicit.

This has been making me wonder if it's necessarily possible to get a coherent definition of "domain name" at all, which is why I was happy to see the definition Lyman proposed. 

To answer Warren's question a few days ago, I think that if we can get general agreement on a definition, even one so abstract that it doesn't tell us very much about how to do anything immediately useful, we can use it to reason about the attributes the domain name space actually has-- and what attributes people might think it has, perhaps incorrectly or inconsistently. 

In my dreams this leads to some notions about what kinds of protocols could reasonably re-use domain name abstractions for naming, when we might actually want to stick to DNS as the wire protocol, and when we might want to do something entirely different. See Ed's domain-name draft and:

https://datatracker.ietf.org/doc/draft-hardie-resolution-contexts/
https://www.ietf.org/mailman/listinfo/arcing


Suzanne