Re: [Cbor] Adam Roach's Discuss on draft-ietf-cbor-cddl-06: (with DISCUSS and COMMENT)

[Jeffrey Yasskin <[e-mail]>]

On Tue, Nov 20, 2018 at 10:56 AM Michael Richardson <[e-mail]>
wrote:

>
> Carsten Bormann <[e-mail]> wrote:
>     Adam> With the lack of any version indicators in CDDL, this seems like
> a
>     Adam> straight-up interoperability issue.
>
>     >> Carsten Bormann <[e-mail]> wrote:
>     >>> There is no expectation that a tool that does not implement a
> control
>     >>> operator is particularly useful with a specification that does use
> it.
>     >>
>     >> okay, but wouldn't it still be good if there was a declaration at
> the top of
>     >> the file as to which revision of the specification the CDDL is
> assuming,
>     >> just so that the tool can fail gracefully?
>
>     > Once there is such a revision, that might make sense.
>
> Not really. It needs to in this version.
> Something like:
>           "CDDL RFC9123"
>
> at the top....
>
>     > Adding a control operator is not a revision — that is the whole point
>     > about exposing an extension point.
>
> Sure.
>
>     > (A tool can easily compare its inventory of control operators with
> the
>     > set used in the specification; no need for a separate declaration
>     > here.)
>
> So, the error is:
>     "RFC9756 control operator found in specification that claims RFC9123"
>

I think Carsten's point is that in order to declare the set of control
operators in use, it's not sufficient to declare just the CDDL RFC. A CDDL
document might be using CDDL RFC9123 and also control operators from
RFC9234, ISO123456, and W3C REC better-cddl-1.

Needing to list all of those, or even just explicitly list the control
operators in use, seems like a pain, and doesn't seem to help old tools
reject new controls much earlier in processing.

Jeffrey