[Cbor] Re: dCBOR Rust Implementation Now Supports Bignums

[Carsten Bormann <[e-mail]>]

On 2025-12-20, at 14:05, Anders Rundgren <[e-mail]> wrote:
> 
> On 2025-12-20 13:48, Rohan Mahy wrote:
>> If we had a time machine and could change the spec before it was implemented, I would have the spec do what Wolf described. However this is implemented in enough places that changing this aspect of preferred serialization is going to break something.
> 
> I (of course) don't know for sure, but given the fact that Bignums have limited utility

In your world.

> and are represented in another way(!) in basic serialization,

Huh?

> I don't think a change would break many *applications*.  

This might be because your field of view is a bit narrow^W focused.

> Breaking *implementations* is less of a problem.  

Not all implementers will agree.
(It is a big problem if it is not clear whether a “fix” is really just a fix or another section of the smorgasbord.)

> As I wrote before, the CDE way is not even normative (4.2.2) using IETF nomenclature and practices.

Please occasionally reread RFC 8949:

3.4.  Tagging of Items

[...] The tag definition may specify a preferred
   serialization (Section 4.1) that is recommended for generic encoders;
   this may prefer basic generic data model representations over ones
   that employ a tag.

3.4.3.  Bignums

[...] preferred serialization (Section 4.1) never makes use of
   bignums that also can be expressed as basic integers (see below).

   [Most of 3.4.3 is about the consequences of the equivalence defined
   for tag 2/3.]

> A change would affect EDN as well since Bignums would need a different syntax, like the n-suffix in JavaScript.

Here the idea of having a single tag that can be slapped around any RFC 8949 integer would shine.

(If we don’t do that, an application-specific extension is the way to go.)

Grüße, Carsten