On Wed, Dec 11, 2024 at 11:44:47AM -0600, Orie Steele wrote:
> Right.
> 
> I think the main thing we lose from going from 3 numbers to 1 is
> parsability.
> If the algorithms are meant to be treated as opaque strings / numbers and
> never parsed, then the single number approach (aligned with cose
> registration requests) seems like the best path forward to me.

While it would be very nasty hack, the COSE numeric algorithm space is
large enough to embed the entire HPKE algorithm combination space (in
9 byte range). 

I think some very hacky test code I wrote (not sure if it ever actually
wroked) stuffed HPKE algs into 0x4845000100010001 to 0x4845FFFFFFFFFFFE
range (4845 => HE).


Both draft-ietf-jose-hpke-encrypt and draft-ietf-cose-hpke need to have
concept of COSE-HPKE/JOSE-HPKE algorithm disjoint from any concrete
algorithm registrations those drafts make.

This is to avoid later drafts adding more algorithms from having to
effectively re-define COSE-HPKE/JOSE-HPKE (or leave a lot of interop-
critical stuff implicit). Not only is this overly tedious, it is also
very error-prone[1].


[1] E.g, draft-ietf-jose-hpke-encrypt partially re-defines JWE Key
Encryption, and gets it wrong.




-Ilari

_______________________________________________
jose mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to