Hi Brian!

Thanks for raising these issues. I have created a PR to reflect the feedback:

https://github.com/cose-wg/draft-ietf-cose-hpke/pull/160/changes

I also added the note about the functionality of some HPKE libraries regarding the info and aad input.

Ciao
Hannes


Am 21.07.2026 um 12:11 schrieb Brian Sipos:
All,
I was able to work-around the API limitation and was able to successfully decrypt (yay!) the example from Section 5.1. Unfortunately, the plaintext output that I got has an additional byte at the end (as ASCII it is a newline). It may be best for the document to include the hex form of plaintext as h'546869732069732074686520636f6e74656e742e*0a*' to avoid any possible ambiguity.

Brian S.

On Tue, Jul 21, 2026 at 2:37 AM Ilari Liusvaara <[email protected]> wrote:

    On Mon, Jul 20, 2026 at 09:09:14PM -0400, Brian Sipos wrote:
    > Ilari,
    > Thank you, I now better understand the relationship to RFC 9180.
    It might
    > be helpful to give an aside somewhere (intro section?) to
    explain that the
    > mechanism of this document is not compatible with strict minimal
    RFC 9180
    > implementations and requires the API of the bis document.

    It is compatible. RFC 9180 talks about using both, which COSE HPKE
    does
    not do (unless using SuppPrivInfo, which is non-interoperable anyway).
    The KE stuff uses info, but that is on layer >0, so should be small.

    The problem is libraries that do not implement single-shot aad at all,
    which is something RFC 9180 does not talk about. However, the text
    could be interpretted to encourage not supporting aad for single-shot.

    Note that if library supports multi-shot, you can emulate single-shot
    by creating a context and then using it once (since that is how
    single-
    shot internally works).


    > I will continue to track this with Py cryptography library
    authors for
    > eventual compatibility with the 9180bis spec. Do you have a
    feeling on the
    > stability of the current version of that draft?

    Where the specifications overlap (and not hitting any implementation
    limits), the behavior is fully compatible.

    - RFC9180bis has higher minimum limits (e.g., on info size).
    - RFC9180 has stuff RFC9180bis does not (e.g., auth modes).
    - RFC9180bis has stuff RFC9180 does not (e.g., single-step KDFs).

    I suppose one should be careful with things like single-step KDF with
    auth modes (I have written HPKE library that somehow supports that).




    -Ilari

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


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

Reply via email to