RE: [saml-dev] holder-of-key subject confirmation

["Scott Cantor" <cantor.2@osu.edu>]

> Okay, but why can't I draw the same conclusion about the name?

I'm trying to say that you can, but that you:

- can't do so *safely* absent other constraints
- *can* do so EVEN IF the KeyInfo contains a certificate as well

In other words, SAML does NOT profile KeyInfo, never has, and until we do
(for some purpose, e.g. metadata), you're free to do anything you can
sensibly argue to your deployment base makes sense, but should think it
through.

When I said, snarkily, not in my book, I was referring to me as a deployer,
essentially.

> If <KeyInfo> contains a key, the RP confirms the subject if the presenter
> proves possession of the key.  If <KeyInfo> contains a name, the RP
> confirms the subject if the presenter proves itself to be the named
> subject.

Not true. In both cases you must prove possession of a key. The difference
is in how the key is identified by the IdP, and that is simply an unprofiled
hook. But in most cases I've seen, using subject name is interpreted to mean
"presents a certificate from a trusted source containing that name". Which
is why I'm saying "sure, you can do that, but you absolutely need the
additional constraint of what sources are trusted".

As an example, the language you use above implies I could satisfy holder of
key by logging into the RP with a password that matches an account with that
subject name. I would claim that's a pretty clear distortion of the intent
behind HoK.

-- Scott