Re: [kitten] Fw: New Version Notification for draft-mills-kitten-sasl-oauth-02

[Sam Hartman <[e-mail]>]

>>>>> "Cantor," == Cantor, Scott E <[e-mail]> writes:

    >> I think all you have to do is communicate the channel binding
    >> type you support.

    Cantor,> I think you mean "the type you used", at least based on my
    Cantor,> reading of GS2.

I do.

    Cantor,> On the general subject of tls-unique vs. endpoint, I had a
    Cantor,> brief conversation with Simon about it, and I think some
    Cantor,> clarification on how server applications would be expected
    Cantor,> to use tls-endpoint would be useful.

Probably.

    Cantor,> How is it expected that servers would distinguish clients

I'd expect that is what the SASL authentication is for.
(Note that HTTP is more complex; you need cookies or some other
distinguisher there).

In something like IMAP, at the end of SASL, we have:

1) Client did mutual auth with server. It knows that the server at the
other end of the TLS exchange with a given certificate is the party
named in the SASL exchange.
It knows that a particular TLS connection reaches that party.

2) Server knows that  the client is as named in the SASL exchange.
Server also knows that the client believes that the party identified by
the particular endpoint CB is the only party besides the client involved
in the TLS exchange.

One critical part of #2 is that if the client didn't confirm the channel
binding, the client would have given up rather than starting  the SASL
authentication.


Unless I'm missing something (wouldn't be the first time) I think 1 and
2 are sufficient for a non-HTTP use case.