[Rats] Re: I-D Action: draft-ietf-rats-ar4si-10.txt

[Thomas Fossati <[e-mail]>]

On Mon, 18 May 2026 at 17:54, Mandyam, Giridhar
<[e-mail]> wrote:
>
> AMD General
>
> > Nothing major, but it's a feature that some EAR users (Confidential Containers/Trustee) plan to use when merging trustworthiness vectors from chained policies.
>
> Thanks for responding.  I hope these EAR users are not using AR4SI for SEV-SNP trustworthiness assessment, because as I mentioned in my last review of the document (https://github.com/ietf-rats-wg/draft-ietf-rats-ar4si/issues/42) this document makes an assertion on mandatory identity requirements (Table 1) that are not considered mandatory in SNP, and do not result in any compromise of the confidential compute guarantees offered by an SNP-enabled platform if the identifier is not present.

They support SNP, but, as far as I am aware, they do not use the
secure interaction model of AR4SI. They only reuse the semantics of
the trustworthiness vector.
So, they simply disregard any identity requirements imposed by the
AR4SI secure interaction model.
That said, you should probably check my non-authoritative assertions
with them directly :-)

cheers, t