Re: [dispatch] Tiny update to RFC 3405

[Timothy Mcsweeney <[e-mail]>]

I'll take that as a yes.

On July 6, 2020 at 6:13 PM Ted Hardie <[e-mail]> wrote:

Howdy,

On Mon, Jul 6, 2020 at 1:27 PM Timothy Mcsweeney < [e-mail]> wrote:

Ted,

Do you not want my scheme in the arpa zone?

I'm just trying to avoid a silly state, where we say do X but don't let anyone do X.  If this is published, then we'll have a clear rule that matches the current system.  But if the consensus turns out to be that the IETF tree should return for the sole purpose of registering schemes going into .arpa, I would go along with that as well. 

I don't support allowing provisionals in the zone because they are first-come-first-served as of RFC 7595 and some of the registrations are very vendor specific (for example, many of the ms- uri schemes in the provisional category at https://www.iana.org/assignments/uri-schemes/uri-schemes.xml" rel="nofollow">https://www.iana.org/assignments/uri-schemes/uri-schemes.xml are Microsoft-only).  That's not in-line with the purpose of .arpa as an infrastructure domain.

If you disagree, please write up your proposal.  I trust DISPATCH will ensure that it gets discussed in the same place as this proposal.

regards,

Ted

On July 6, 2020 at 4:19 PM Ted Hardie < [e-mail]> wrote:

On Mon, Jul 6, 2020 at 12:09 PM Timothy Mcsweeney < [e-mail]> wrote:

Hi,

>If I understand you correctly, you are arguing that because URI.arpa was not >depopulated when the IETF tree was dropped, registrations can still be made >according to the old rules as if there still were an IETF tree.

I'm arguing that, as it sits right now, in order to insert a record into uri.arpa, 

you have to have a scheme name registered.  

RFC 3405 is pretty restrictive in its language:

3.1 URI.ARPA Registration

3.1.1 Only Schemes in the IETF Tree Allowed

   In order to be inserted into the URI.ARPA zone, the subsequent URI
   scheme MUST be registered under the IETF URI tree.  The requirements
   for this tree are specified in [10].

Given the "Only" and the RFC 2119 "MUST", I don't think a plain reading of the text supports the view that any URI registration is sufficient.   Section 3.1.2 also reinforces that the registration must be prior and then the record insertion must pass IESG review; that section does not given the IESG the right to waive the requirements:

3.1.2 Scheme Registration Takes Precedence

   The registration of a NAPTR record for a URI scheme MUST NOT precede
   proper registration of that scheme and publication of a stable
   specification in accordance with [10].  The IESG or its designated
   expert will review the request for

      1.  correctness and technical soundness

      2.  consistency with the published URI specification, and

      3.  to ensure that the NAPTR record for a DNS-based URI does not
          delegate resolution of the URI to a party other than the
          holder of the DNS name.  This last rule is to insure that a
          given URI's resolution hint doesn't hijack (inadvertently or
          otherwise) network traffic for a given domain.

regards,

Ted Hardie

On July 6, 2020 at 2:51 PM Ted Hardie < [e-mail]> wrote:

Howdy,

On Mon, Jul 6, 2020 at 11:26 AM Timothy Mcsweeney < [e-mail]> wrote:

Ted, 

>
>Yes, this came up because of a proposed registration. Since it was yours,
>perhaps you would like to provide the link to the group?

There is already a mailing list for that.

>Once DISPATCH decides where this ought to be discussed, we can discuss that
>outcome or the update to BCP 35 to restore the category as alternatives.

There should not be a discussion at all. 
1.  Section 5 of RFC3405 isn't broken.  Maybe you were confusing it with
     Section 5 or RFC4395? 

As I note in the extremely short document:

   The document requires that registrations be in the "IETF
   tree" of URI registrations.  The use of URI scheme name trees was
   defined in RFC 2717 [RFC2717] but discontinued by RFC 4395 [RFC4395].
   Since the use of trees was discontinued, there is no way in the
   current process set out in BCP 35 [RFC7595] to meet the requirement.

If we leave things as they are, no registrations can be made, because the category is gone.  We can change it to require permanent registrations instead (as this document suggests) or you could propose something different (e.g. updating BCP 35 to recreate the IETF tree for these registrations).

2. Regardless, any discussions should really wait until after upcoming
registrations or appeals of those registrations, or appeals of those appeals are
completed.

>The current rules cannot work because they reference a category that no
>longer exists. To put this differently, if they don't change, there can be
>no more registrations in URI.arpa.

The current rules are working just fine. 
HTTP, among others, are still in the uri.arpa zone proving that the RFC3405
Section 3.1.1 reference [10] lives on through the obsoleted RFCs to the current
spec and can be seen in totality in IANA's list of URIs.

If I understand you correctly, you are arguing that because URI.arpa was not depopulated when the IETF tree was dropped, registrations can still be made according to the old rules as if there still were an IETF tree. 

That's not how the IETF process treats obsoleting BCPs; see the IESG statement at https://www.ietf.org/about/groups/iesg/statements/designating-rfcs-historic-2014-07-20/" rel="noopener nofollow">https://www.ietf.org/about/groups/iesg/statements/designating-rfcs-historic-2014-07-20/.

This situation has pointed out that there was a bug introduced by RFC 4395 that was carried forward into RFC 7595, because they did not address the dependency on the removed IETF tree in BCP 65.  This document is one way to address that bug.  If you wish to suggest others, that's fine, but we still need DISPATCH to identify where the discussion should happen. 

regards,

Ted

On July 6, 2020 at 12:15 PM Ted Hardie < [e-mail]> wrote:

Howdy,

On Sat, Jul 4, 2020 at 2:28 AM Timothy Mcsweeney < [e-mail]> wrote:

Ted, 

In your opening email to the 400 highly respectable people on this list you say:

"As it happens, there are very few registrations in URI.ARPA, so we did not catch it and fix it before now."  

How did you "catch it"? 

Was there a pending registration? 

Is there still a pending registration? 

Yes, this came up because of a proposed registration.  Since it was yours, perhaps you would like to provide the link to the group?

It would really be bad to try to change the rules while something was pending.

The current rules cannot work because they reference a category that no longer exists. To put this differently, if they don't change, there can be no more registrations in URI.arpa. 

Once DISPATCH decides where this ought to be discussed, we can discuss that outcome or the update to BCP 35 to restore the category as alternatives.

regards,

Ted Hardie

I can't speak for the others but some of them might want to know why after almost 20 years of there being zero problems with RFC3405 it suddenly needs to get "fixed".

_______________________________________________
dispatch mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/dispatch" rel="noopener nofollow">https://www.ietf.org/mailman/listinfo/dispatch