On Thu, Jul 20, 2023 at 08:34:15PM +0100, Stephen Farrell wrote: > > > On 20/07/2023 20:27, Ilari Liusvaara wrote: > > On Thu, Jul 20, 2023 at 08:07:19PM +0100, Stephen Farrell wrote: > > > > > > > > > I'd further encourage minimising the number of values to be > > > defined now to only those that are needed now. And even > > > further encourage general libraries to support only a tiny > > > number of the options that'll inevitably get defined. Two > > > or maybe three would probably be plenty to start from. > > > > Unfortunately, even currently it would be 12. > > > > ... Then there is I-D that would push it to 18... > > > > ... And foreseeable extension, that would push it to 27... > > > > ... And the rest I can't even estimate. > > IMO, that means someone is doing things wrong. But yeah they > probably won't agree;-(
I think the only thing wrong there is use of ciphersuites. :-) > > Basically, it easier to just support all the combos than to support some > > fixed list of ciphersuites. > > And from experience implementing HPKE: The various non-base modes are > > much more annoying than all the KEM/KDF/AEAD combos. > > 90 suites and 4 modes is a silly amount of combinations, yes, > esp since afaik all protocols so far only use base mode etc. It is currently only 54 "ciphersuites". Oh, except only 45 of those work with auth/authPSK modes... And there is a I-D in pipeline that would increase the number of "ciphersuites" to 135... I think HPKE should have combined KEM and KDF, since the KEM turns out to have builtin KDF anyway, but oh well. That would have cut the number of current ciphersuites to 18 (45 after that draft). -Ilari _______________________________________________ COSE mailing list [email protected] https://www.ietf.org/mailman/listinfo/cose
