Re: What to reference when importing an IANA module?

Hi -

On 2023-01-13 10:20 AM, Kent Watsen wrote:
> 
> 
>> On Jan 13, 2023, at 11:25 AM, Benoit Claise <benoit.claise=40huawei.com-Tr9gZwTxerDR74oF6e/[email protected]> wrote:
>>
>> Hi Tom,
>>> Yes I do think that people outside the IETF may be ignorant of the nuances of the way the IETF works and  may not realise that a URL to the IANA website must be used in preference to an RFC.  There is more to YANG modules than extracting the code from somewhere in order to incorporate it into something.  I have even seen RFC reference the obsolete list of possibilities  in the RFC that set up an IANA registry
>> If this is the case (And Randy supports this), then we should update RFC 8047.

Benoit's reference to RFC 8047 had me puzzled until I saw Kent's
response regarding RFC 8407.  :-)

> Agreed - as a hold for document update?
> 
> Currently RFC 8407, Section 3.9 says:
> 
>     For every import or include statement that appears in a module
>     contained in the specification that identifies a module in a separate
>     document, a corresponding normative reference to that document MUST
>     appear in the Normative References section.  The reference MUST
>     correspond to the specific module version actually used within the
>     specification.

I agree with Kent's "hold for document update" assessment.  The
difficultly with the existing text is that it correctly reflects
the concerns that were at the forefront when it was written -
e.g. making it as easy as possible for developers to get the
necessary context for implementing a module, but, as far as
I can recall, the group hadn't thought as deeply about
registries spun off from an initial document.

> Want to take a swing at it?

Not me.  :-)  There are competing requirements, and the "best" answer
will very much depend on each situation.  I think the *spirit* of
the RFC 8407 Section 3.9 is  "point to whatever resource will be most
enlightening to the developer / user."  But the letter of the law is
"point to whatever is needed to generate a tree of normative
reference dependencies" - that is, use what will be most helpful
to the people writing the standards.  There's a point to both kinds of
pointers.

When in doubt, my preference is to go whichever way will make it
harder for implementations to get it wrong.

Randy