[Cbor] Invariants (Re: packed review (was: Re: Reminder and call for agenda: CBOR WG Virtual Meeting on 2022-06-01))

[Carsten Bormann <[e-mail]>]

I think we are discussing an extension point for cbor-packed, the function tag.

Any proposal for this extension point will need to define what is invariant over the time this extension point is exercised more and more.

What seems very important for me is that new function tags can be defined in such a way that an unpacker can find out whether something is being used as a function tag or not.

The current proposal does this by giving the function tag role exactly to the dominating side of the packed reference tag (type-0: argument, type-1: reference tag content = rump).

This is a bit hard to visualize sometimes because there are always two tags: The (non-extensible) reference tag, and the function tag pointed to by the reference tag.  By making reference tags non-extensible, we make sure that all implementations know when these are active.

This approach does not allow to freely evaluate tags occurring anywhere.
That is hard to make extensible in such a way that false interoperability can be detected.
It also probably requires a more complex processing model.

http://notes.ietf.org/notes-cbor-not-interim-2022-06-29

Grüße, Carsten