Hi Paul,

Thanks for the review!

We just submitted -16 addressing your comments, see responses below.

We agree essentially on everything, details in this pull request:
https://github.com/cose-wg/CBOR-certificates/pull/302

Thanks,
Göran


From: Paul Wouters <[email protected]>
Date: Thursday, 8 January 2026 at 21:55
To: [email protected] 
<[email protected]>
Cc: cose <[email protected]>
Subject: AD review draft-ietf-cose-cbor-encoded-cert-15

Hi,

I've done my AD review of draft-ietf-cose-cbor-encoded-cert-15

I have a few comments and questions, but I think the most important item is 
whether we asked the TLS WG what they think about making the C509 TLS 
Certificate Types RECOMMENDED Y to implement. It seems quite a lift to require 
all TLS servers would implement C509 at this point in time? The field is 
documented to mean:

If the "Recommended" column is set to "N", it does not necessarily
mean that it is flawed; rather, it indicates that the item either
has not been through the IETF consensus process, has limited
applicability, or is intended only for specific use cases.

Which I would argue matches better with RECOMMENDED N

[GS] Agree

See the rest of my AD review below,

Paul


The Abstract is large and the Introduction repeats many things. Can we
deduplicate this and make the Abstract smaller and more consise?

Please add a sentence: This document updates RFC 6698 in the Abstract.

[GS] The Abstract is updated and now includes this text.


        validityNotAfter: ~time / null,

Why can this be null? I thought X.509 certs always have an end date?
(reading further, this is explained in 3.1.5)

[GS] Right, the parameters are explained in turn in the following subsections. 
No change made.

        the octets 0xfe and 0xfd are used instead of 0x02 and 0x03
        in the CBOR encoding to represent even and odd y-coordinate,
        respectively.

I would prefer to rewrite this without using the "respectively" construct
to help non-native english speakers.

[GS] Done, see PR.

This does not parse:

        using of the reserved CBOR tag

I assume "of" just needs to be removed?

[GS] Correct, fixed here and in other locations using the same phrase.


        An implementation MAY only support c509CertificateRequestType = 0.

This is an odd way of saying, I think, that c509CertificateRequestType = 0 is 
mandatory to implement.

[GS] This is a remnant from when there were fewer options / values. Rewritten 
as "An implementation MAY only support certain values of 
c509CertificateRequestType."

There seems to be a spelling error in: TBSCertificatdsveRequest
Should this be: TBSCertificateRequest ?

[GS] Correct, fixed.

        For example, if the EST server includes a subjectAltName with
        a partially filled extensionValue, such as iPAddress with an
        empty byte string, this means that the client SHOULD fill in
        the corresponding GeneralName value

This makes it appeat the EST server can put in "10.0.2" and the client can only
pick that range, or "example.com<http://example.com/>" and the client can pick 
only subdomains from
"example.com<http://example.com/>", but it is unlikely that you meant this? 
Perhaps rephrasing this
to avoid this interpretation?

[GS] Thanks, we made the example more explicit in the PR.


"Brotli [RFC7932]" is not clarified until after "Brotli" has been mentioned a 
few times already.
Please move the "[RFC7932]" clarification to the first instance of the use of 
"Brotli".

[GS] Done.

This sentence is unreadable, please rewrite:

        Any issue to decode or parse a C509 certificate should be handled
        by the certificate using system as would the issue of parsing
        the corresponding X.509 certificate.

Perhaps:

        Any problems with decoding or parsing a C509 certificate should be 
handled
        exactly as how such errors would be handled for the corresponding X.509 
certificate.

[GS] Done.

Section 9.5, should the text "Name and Identifiers are informal descriptions" 
maybe be included
as a note above the table?

[GS] Done.

Why is Postal Code defined, but not Postal Address? Is that because Street 
Address is used?

[GS] This was discussed in
https://github.com/cose-wg/CBOR-certificates/issues/255
and it was decided to not specify it in CBOR since we didn’t see much room for 
simplification.

Section 9.10. C509 Extended Key Usages Registry

A good list of EKUs is:
https://docs.digicert.com/zf/trust-lifecycle-manager/define-policies-to-ensure-compliance/certificate-attributes-and-extensions/extended-key-usage.html

I happen to know (as an IPsec/IKE expert) that some implementations (like 
Microsoft) requires
EKUs for IPsec which are not defined here.

I guess these can be added later.

[GS] Yes, they can easily be added later.


Section  9.17. TLS Certificate Types Registry

This sets the new type to RECOMMENDED. Since this is (currently) very much an 
IoT related document, I am not
sure if it qualifies as something to be set to RECOMMENDED Y, meaning all 
servers/clients MUST/SHOULD implement
this. Has there been a conversation with the TLS WG or the Designated Experts 
for this Registry?

[GS] As proposed above, we agree to change this to RECOMMENDED N.



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

Reply via email to