The -03 draft seems fine. :-) One comment below.
LL > On Jan 18, 2023, at 6:46 AM, Russ Housley <[email protected]> wrote: > > Laurence: > > Thanks for the review. Your comments have made the document better, even > though I am pushing back on one of them. > >> Here’s a few comments that are primarily about wording and presentation. >> The actual protocol bits defined here seems ready to me. > >> 2) Section 4 >> If it were me, I’d move / remove many of the paragraphs in section 4 to >> security considerations. For example the stuff about re-using IVs, how easy >> it is to use incorrectly and discussing about forging. My preference would >> be that section 4 only describe bytes on the wire, protocol and operations. > > This is an attempt to warn implementers early and often. Perhaps the balance > is wrong. If so, I'd like to hear from others. It takes a lot of energy to absorb exactly what an implementation is supposed to do and not do from an RFC under schedule and resource pressure. I know this first hand from creating commercial implementations of X.509, CMS, CBOR and COSE. The details involved in writing code (e.g, bits on the wire) are very different from the end-end system architecture (selecting a algorithm or keying scheme). Then sometimes design rationale gets in the mix too. I like standards best that keep these three separate: 1) When I’m trying to figure out whether to use a standard or not, the design rationale is helpful. Put this in the intro sections. 2) When I’m designing the end-end system that the standard goes into, I want to read the security considerations. 3) When I writing the code, I want to just know what bits on the wire are required. RFC 9052 is fairly good about this. It’s true that some security considerations may have value when writing the code, but it’s often not the case and is not the case here. I’m not requesting any changes, just providing background for my comment. LL
_______________________________________________ COSE mailing list [email protected] https://www.ietf.org/mailman/listinfo/cose
