[Cbor] Re: 🔔 WGLC on draft-ietf-cbor-cde-08

[Anders Rundgren <[e-mail]>]

On 2025-03-18 18:47, Carsten Bormann wrote:
> On 2025-03-18, at 18:03, Anders Rundgren <[e-mail]> wrote:
>>
>> In RFC 8949 4.2.2 Rule 2 is a "friendly advice".  I.e. non-normative.
> 
> That numbered list is preceded by:
> 
>        Example rules for this are:
> 
> Please read.
> 
> The thing we didn’t have when we wrote this is a strong terminology distinguishing deterministic encoding from deterministic representation.  That led to the phrase "protocol's deterministic encoding” in the sentence leading up to this (what these are examples for), which in today’s terminology would be “application protocol’s ALDR rules”.
> 
> What you didn’t quote was the text in 4.2.1:
> 
>        Floating-point values also MUST use the shortest form that
>        preserves the value, e.g., 1.5 is encoded as 0xf93e00 (binary16)
>        and 1000000.5 as 0xfa49742408 (binary32)
> 
>> In draft-ietf-cbor-cde-08 the same text is "stronger" but still non-normative.  No SHOULD or MUST in sight.
> 
> Not needed, see above.
> 
>> That is, CDE effectively MUST be accompanied by a derived document in order to provide a normative deterministically encoding standard.
> 
> No.
> 
>> Since dCBOR is the only derived CDE document, dCBOR effectively becomes the norm.
> 
> No, not at all.
> 
>> Not that I actually care about CDE, but this situation appears pretty absurd seen from my watchtower.
> 
> It sure is absurd to come to this misinterpretation.

Well, the line

"2. If floating-point numbers are emitted"

is a bit of a riddle.  I guess this is where deterministic encoding and representation MAY differ.

Anders

I'm confident that the intended audience ("market" as I say) for U-CBOR compatible tools, won't ever have to bother about such issues.  Through state-of-the-art, OO API concepts and encapsulation, this lot should be equally unaware of ALDR as well.  In its latest incarceration, even the media type question was reduced to zero, as is already the case for comparable systems using JSON, like Open API:
https://www.ietf.org/archive/id/draft-rundgren-universal-cbor-06.html#name-media-type

> 
> Grüße, Carsten
> 
>