Re: [Doh] Use of TXT records

["Hewitt, Rory" <[e-mail]>]

Understood.

Well presumably TXT records (given their name) were ORIGINALLY INTENDED to only store human-readable text. But what defines human-readable? I mean, it's all ASCII characters (no binary data), so it's all "human-readable" to some extent.

The problem is surely that because there is no standard that says "The first X characters in a TXT record must be one of the following IANA-registered types, and the subsequent data must follow the related IANA format", then ANYTHING which reads a TXT record must do some (possibly complex) processing to determine what sort of TXT record it is?

For instance, SPF records just start with v=spf1 followed by a blank, which REALLY bugs me, because there's nothing that defines what 'type' of TXT record it is. But the subsequent text is 'kinda' human-readable - you just need to know its format.

I'm fine with using TXT records for DoH, but I would love them to start with type=DOH followed by a blank. Whatever format beyond that is used doesn't matter, as long as it's consistent. Since it's going to be IP addresses, what's the concern - it wouldn't even need to be Base64-encoded, would it?

If you use a new RRTYPE, then every single document ever written about DNS record types will have to be updated to include the new RRTYPE.

Maybe a little hyperbolic, but only a little...

LSS; Use TXT records.

Rory

From: Ben Schwartz <[e-mail]>
Sent: Friday, March 15, 2019 12:24 PM
To: Hewitt, Rory <[e-mail]>
Cc: Paul Hoffman <[e-mail]>rg>; DoH WG <[e-mail]>
Subject: Re: [Doh] Use of TXT records

On Fri, Mar 15, 2019 at 12:57 PM Hewitt, Rory <[e-mail]<mailto:[e-mail]>> wrote:
Ben,

How has the use of TXT been a point of controversy in the past?

See for example https://mailarchive.ietf.org/arch/msg/tls/b0DQpCStJ4RSlp7uVXSobS-ssjA

Is it a general "TXT records are bad" or "Adding new TXT records is bad" or "TXT records have never been standardized or formatted consistently among their various uses, so there's always the possibility of clashes".

My understanding is that some DNS experts believe that TXT records should be reserved for human-readable comments, while others believe that it is more expedient to store data in TXT records than to define a new RRTYPE for each new usage.

Because I, for one, would agree with the last.

Rory

From: Ben Schwartz <bemasc=[e-mail]<mailto:[e-mail]>>
Sent: Friday, March 15, 2019 9:53 AM
To: Paul Hoffman <[e-mail]<mailto:[e-mail]>>
Cc: DoH WG <[e-mail]<mailto:[e-mail]>>
Subject: Re: [Doh] Use of TXT records

To be clear, I'm not objecting to this choice.  I'm merely noting that use of TXT has been a point of controversy in other discussions at the IETF.  Some groups have decided to proceed with TXT, some have preferred to use different or new RRTYPEs, and some have not yet decided.  I would like the working group to consider this question soon so it doesn't delay consensus later.

On Fri, Mar 15, 2019 at 12:27 PM Paul Hoffman <[e-mail]<mailto:[e-mail]>> wrote:
On Mar 15, 2019, at 6:12 AM, Ben Schwartz <bemasc=[e-mail]..org<mailto:bemasc=[e-mail]..org>> wrote:
> 5. Machine-readable content in a TXT record

This is the first I had heard that people objected to using a TXT record, but I might have missed that part of the discussion. Given that this query is for a newly-created SUDN that is only used for looking up DoH information, there is no chance that the TXT record would be in use for anything else.

Having said that, it would be quite easy for this document to create a new RRtype for this particular query. (I would not want to re-use the URI RRtype because it would take more effort to explain the priority and weight, as well as dealing with the common error of people not put the URI in quotations marks as is required for the URI RRtype.)

--Paul Hoffman