Re: [Cbor] Announcement: draft-mcnally-deterministic-cbor-03

[Joe Hildebrand <[e-mail]>]

Nod, that's fine with me, then.

— 
Joe Hildebrand

> On Aug 7, 2023, at 2:21 PM, Wolf McNally <[e-mail]> wrote:
> 
> Joe,
> 
> Carsten’s explanation comports with my understanding, and is reflected in RFC 8949 §4.2.1 on determinism where it says, "Preferred serialization MUST be used. In particular, this means that arguments (see Section 3) for integers, lengths in major types 2 through 5, and tags MUST be as short as possible…”
> 
> This is reiterated in the dCBOR I-D §2.1: "dCBOR encoders … MUST encode variable-length integers using the shortest form possible.” The I-D also extends this with an explicit validation requirement: "In addition, dCBOR decoders … MUST reject any variable length integers that are not encoded in the shortest form possible."
> 
> https://www.ietf.org/archive/id/draft-mcnally-deterministic-cbor-04.html#section-2.1-3.1
> 
> ~ Wolf
> 
>> On Aug 7, 2023, at 9:12 AM, Carsten Bormann <[e-mail]> wrote:
>> 
>> On 2023-08-07, at 18:00, Joe Hildebrand <[e-mail]> wrote:
>>> 
>>> One more thing, there may need to be another note that says that the integer descriptors for complex types (array, map, tag, byte string, utf8 string) MUST follow the rules for mt 0/1.  It's implied strongly and probably difficult to implement any other way, but 
>>> might be worth mentioning.
>> 
>> Hi Joe,
>> 
>> descriptor is not a word we use in RFC 8949.
>> We have something that is used with all of these, which is by definition an unsigned integer: The argument [1].
>> 
>> The argument can represent any integral number between 0 and 2**64-1.
>> 
>> The argument is directly used as the mt0 value (the mt1 value is -argument-1, or, in C integer arithmetic, ~argument).  For mt2 and mt3, the argument gives the length of the string in bytes.  For mt4, the number of elements in the array.  mt5, number of pairs in the map.  mt6, the tag number.  mt7 is a bit more complicated.
>> 
>> So, by definition, the argument already behaves like an mt0, because mt0 is defined by the argument...
>> 
>> Grüße, Carsten
>> 
>> [1]: https://www.rfc-editor.org/rfc/rfc8949.html#name-specification-of-the-cbor-e
>> 
>