On Thu, Sep 07, 2023 at 07:31:28PM +0000, Michael Jones wrote: > That's useful information, Ilari. Could you send us a pointer to your > rules write-up so we can factor it into the discussion?
https://mailarchive.ietf.org/arch/msg/cose/fu5pI-mN2NWTSr-J3klijF-tUGI/ The general principle is starting from every KEM, assigning the one KDF that makes the most sense, and then assigning the most sensible variant from every family of AEADs that makes sense. Currently there are 6 world-class KEMs and 2 world-class AEAD families, so there would be 6*2=12 ciphersuites. That message also seems to cover what would be needed for ciphersuites. The "sub-registry" stuff is trying to solve the following issue: Say ShangMi (China) KEM/KDF/AEAD gets added to HPKE, and support for it is desired in COSE-HPKE. The spcification of the COSE algorithm (singular!) should be minimal, because: - Duplication makes it harder to write the specification. - The duplicated parts might not actually be quite duplicated. Which in turn causes compatiblity and interoperability issues. Another possible idea would be to name the algorithms systematically. E.g. - HPKE-v1-base-10-1-1 - HPKE-v1-base-10-1-3 - HPKE-v1-base-11-2-2 - HPKE-v1-base-11-2-3 - HPKE-v1-base-12-3-2 - HPKE-v1-base-12-3-3 - HPKE-v1-base-20-1-1 - HPKE-v1-base-20-1-3 - HPKE-v1-base-21-3-2 - HPKE-v1-base-21-3-3 - HPKE-v1-base-30-1-2 - HPKE-v1-base-30-1-3 (And then future extensions) ... Which allows telling from just the name how to use it (the numbers are in order KEM/KDF/AEAD and are in hex without leading zeros). -Ilari _______________________________________________ COSE mailing list [email protected] https://www.ietf.org/mailman/listinfo/cose
