Re: [netconf] [Technical Errata Reported] RFC8341 (6493)

[Andy Bierman <[e-mail]>]

On Fri, Apr 2, 2021 at 8:14 AM Kent Watsen <[e-mail]> wrote:

>
> I agree with Juergen that section 9.13.2 of RFC 7950 applies and hence my
> opinion is that this errata, against RFC 8341, should be rejected.
>
>
> +1 (as contributor)
>
>
> However, I also think that RFC 7950 isn't as clear as it should be.  Given
> that there is an explicit ABNF construct for instance-identifier, I really
> think that this ABNF should explicitly require prefixes for
> node-identifiers, avoiding any potential ambiguity in what is allowed.
>
>
> Should a YANG-next issue be filed?
>


We could think a bit about the real problems with instance-identifier
instead of academic discussions.

The ABNF only describes the syntax of the default-stmt representation of an
instance-identifier.
Since such a default would be extremely rare, this is not an interesting
problem.
The RFC is clear that prefixes are required.

The protocol message (or artifact) representation of an instance-identifier
is a much more relevant use-case.
XML has its own rules for binding namespaces to QNames in the context of an
XML instance document.
These rules are clear, so no problem to be solved here.

The real problem (for yang-next?) is that the instance-identifier and
identityref data types
do not have any context-free representation possible, let alone a canonical
representation.

The solution already exists in RFC 7951 (6.8 and 6.11).
These encoding rules for identityref and instance-identifier should be used
for all
representations, and the default-stmt, The XML and YANG encodings should be
replaced
(by introducing new types and deprecating the old ones).


>
> K.
>
>
Andy