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

[Paul Hoffman <[e-mail]>]

On Sep 21, 2016, at 10:52 PM, Martin Thomson <[e-mail]> wrote:
> 
> On 22 September 2016 at 11:32, Paul Hoffman <[e-mail]> wrote:
>> On Sep 21, 2016, at 6:16 PM, Martin Thomson <[e-mail]> wrote:
>>> 
>>> On 22 September 2016 at 01:12, Paul Hoffman <[e-mail]> wrote:
>>>> 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.
>>> 
>>> I think that would be a mistake.  Pick one and polish it.
>> 
>> Sorry, this just feels wrong. For example, why should a DNS application have to add a JSON processor that it will never use if it wants to get responses in DNS wire format?
> 
> That's where I think we have a disconnect on use cases.  I want to
> make DNS more accessible to application developers.  I think that some
> others want to take advantage of the fact that no middlebox dares to
> break HTTP.  The wire format in antithetical to my goal.  My thesis is
> that if we have the wire format, there's no incentive to do the other
> thing.

We'll disagree on two axes there.

- You should not be able to prevent a community from using the format they want as long as they use HTTP properly, and "properly" is the goal of this document. Some DNS communities need extremely fast processing of the data and no problems with escaping, so JSON will not serve them well.

- You haven't been following the discussions over the years on DNSEXT and DNSOP. The wire format is actually less popular than JSON. If you somehow managed to make dns-over-http be a single output format (which I hope you don't), I believe that the DNS community would prefer JSON.

> 
>>> Why do I hear so much about the ISE with these things?
>> 
>> Because it's an experimental document and no obvious WG to discuss it in.
> 
> The activity on this list leads me to conclude that this is a
> temporary situation.  And "experimental" is not doing this justice; if
> someone wants to deploy this for realsies - and that seems like it has
> some decent chance of happening - then it's worth doing the protocol
> for realsies too.

If you believe that there will be actual deployment of the JSON format in the next year, I'm happy to keep cycling the draft in a WG or as AD-sponsored standards track. To date, I think I'm the only user (and barely that: I rolled my own more kludgy format a few years ago and still mostly use that).

> 
>> If you want to find an AD who wants to sponsor this, great. At that point, it could go on standards track. To me, this is asking the IESG to do much more work than it needs to for a fairly small document.
> 
> I have to disagree about this being small.  Done right, this could be
> a big deal.

Sounds good to me (or at least to my ego).

--Paul Hoffman