Re: [lamps] header-protection test vectors: should they include intermediate CA?

[Daniel Kahn Gillmor <[e-mail]>]

Hi Ryan--

On Thu 2021-05-27 20:08:45 -0400, Ryan Sleevi wrote:
> There is a slight benefit, although in the context of S/MIME and CMS, it's
> far more esoteric, particularly for S/MIME.

Thanks for the detailed breakdown of the processing, update, and
verification model used by Windows.  fwiw, i'm not expecting that any
MUA receiving these messages would even consider updating their stance
toward what roots are trusted just to process the message -- that would
be a strange (and dangerous) tradeoff.

From the sender's perspective, whether the receiving party has a known
set of trusted roots, or a subscription to a service (like "AuthRoot")
that provides just-in-time delivery of system-trusted roots seems
basically indistinguishable to the sender.

So, I don't see anything in your description that gives me any reason to
think i should ship any root certificates in the PKCS#7 bundle.  The
draft won't do that unless someone else on the list presents a strong
counterargument for doing so.

Given that we don't know who the relying party is going to be, though,
and it's not always clear even to the microsoft CDN how to find a
particular intermediate cert, i don't think this analysis lands on the
same conclusion for shipping the certificate of an intermediate CA,
which is what i was proposing.

So far, it looks like the next draft (-06) *will* go ahead and ship the
cross-signed intermediate cert, so that any receiving party that is
already willing to rely on either of the two example root CAs can
validate the message without needing to chase down any intermediate
cert.

I hope folks will speak up here on the list if they think this is a bad
idea!

        --dkg