Re: [Cbor] Simple values 24..31 (was Re: List of not-well-formed CBOR and test vectors)

[Laurence Lundblade <[e-mail]>]

> On Jul 31, 2019, at 11:52 PM, Carsten Bormann <[e-mail]> wrote:
> 
> On Aug 1, 2019, at 05:01, Laurence Lundblade <[e-mail]> wrote:
>> 
>> Why wouldn’t 24 through 255 be considered unassigned and encoded 0xf8 0x18 through 0xf8 0xff?
> 
> Because we considered those values confusing, and decided to give up on 8 potential values in exchange for reducing this confusion.  See Section 2.3 of RFC 7049 (parenthesis in first paragraph on page 13).
> 
> Whether that was a particularly wise decision can be discussed.  Since we already have to check for too-low values in the second byte of 7.24, checking for 32 is not more onerous than checking for 24 (see pseudocode).  Giving up 8 values seems to be of limited concern given the number of allocations of Simple values so far.  Changing this now, however, would be a backward compatibility issue that we need to avoid.
> 
> Grüße, Carsten
> 

I see the logic, though I kind of think the opposite. Carving out a chunk of simple values as never-to-be-used is also confusing. Confused me :-). Another reason it is confusing is that it makes them (just slightly) different from low-value integers.

Probably a lot of decoders today cannot decode these simple values, so allowing them now would break backwards compatibility.

I think they should be labeled different than “reserved” because that implies they are going to be used at a later time or for some special purpose. That’s why I went looking at extension points and IANA. Here’s a few suggestions:
Never to be used
Disallowed
Forbidden

Agreed that implementation cost is insignificant for any choice here.

LL