Re: [dnsoverhttp] New draft: draft-hoffman-dns-over-http-00.txt

[Paul Hoffman <[e-mail]>]

On Sep 20, 2016, at 7:21 PM, Martin Thomson <[e-mail]> wrote:
> 1. caching is not discussed, seems like it might relate to TTL somehow

We certainly meant to: it's an important part of HTTP. In the section on responses, we say:

In the HTTP responses, the HTTP cache headers are set to shortest DNS TTL in the response.
The HTTP responses SHOULD set the HTTP ETag headers as described in {{RFC7232}}.

Suggestions of what more we should say are greatly encouraged.

> 2. .invalid might be a poor choise:
> https://tools.ietf.org/html/rfc6761#section-6.4

Very good point. If we keep with the "send this special name" scheme, we'll pick another.

> 3. do we really need "id"?  adding extra correlators invites all sorts of crazy

One of our stated goals was to allow any real DNS query. Although it would be bad practice, some servers might treat certain IDs as special.

> The format is important: we should not just bungle a DNS message in a
> response, for example, additionalRRs is an area that might be
> addressed with server push.  

Joe and I purposely kept this document agnostic to the response format. There will be different formats for different use cases, as Shane's draft shows.

> The JSON format is somewhat in the right
> direction, but it needs a lot of work (this is my go native comment,
> consistent with your stated goal to be native).

As a side note, I'm in an informal "last call" for draft-hoffman-dns-in-json with the intention of sending it to the ISE for Experimental status in a few weeks. There is some traffic about that on the DNSOP WG mailing list, and I'm happy to take offline comments as well. I'll be putting out a revised version in the next few days.

--Paul Hoffman