Certification declaration parsing is future-hostile

[bzbarsky-apple]

Problem

The code added in #10825 requires that there be no tags after kTag_DACOriginVendorId. That means future spec versions will not be able to add anything there.

Why are we doing that?

Proposed Solution

Stop requiring CHIP_END_OF_TLV there. Allow CHIP_NO_ERROR as well, and probably CHIP_ERROR_UNEXPECTED_TLV_ELEMENT.

[emargolis]

I don't think it is a good idea to allow TLV elements in the CD, which cannot be interpreted by the code. This is why the code is written this way, which doesn't allow unrecognized elements. Any future formats CD formats will require new SKD version, which knows to interpret new format.

The spec doesn't clearly define format_version element. It looks that we can claim that this parser only supports CD format_version == 1. For any future format_version the parser should be updated.

@tcarmelveilleux ^^

[bzbarsky-apple]

So "new" certification declarations are guaranteed to be be backwards-incompatible with downrev commissioners?

If that's what we are doing, why are we bothering with TLV at all? We could have a more compact fixed-field encoding here....

[bzbarsky-apple]

In either case, if the expectation is that no other tags are allowed, the spec needs to explicitly call that out. That is NOT the general expectation for TLV things.

[woody-apple]

This looks resolved?

[bzbarsky-apple]

Looks like #11838 fixed this.