On Tue, Sep 6, 2011 at 6:19 PM, Sam Hartman
<[email protected]> wrote:
>    Nico> c) Add some AS-REQ flags so that the AS-REP will contain an
>    Nico> pre-auth element that contains a KRB-CRED in it with the
>    Nico> desired tickets.  In this case the enc-part of the AS-REP will
>    Nico> be useless and will be thrown out.
>
>    Nico> d) Add some AS-REQ flags so that the AS-REP's pre-auth method
>    Nico> data will contain an embedded KRB-CRED (in the case of GSS
>    Nico> pre-auth probably using the Null enctype and embedded in a
>    Nico> wrap token).
>
> I don't see the difference between C and D.

I do.

In (c) the RP has to collect the additional PA and bundle it in.  In
(d) the PA method at the KDC end-point takes care of that.  Less
abstraction violation in the RP -> better.

> Also, this all seems like it focuses on the part of the problem where I
> see lots of options rather than the part of the problem I don't know how
> to deal with.
> Namely, I don't understand how to make the key hierarchy work out right
> across the gss-preauth boundary without either
>
> 1) building Kerberos knowledge into the initiator
>
> 2) having the acceptor gain the TGT key.

The initiator MUST have a Kerberos client, else it MUST use a trusted
proxy wherever Kerberos is required.  I've no interest in building a
trusted proxy here.

So the question really is: what kind of violence to the initiator
might we accept?

Now, imagine for a second a variation on GSS_Init_sec_context()
specifically for IAKERB-like mechanisms that outputs...  a credential
handle, just like GSS_Accept_sec_context does...  The KRB-CRED
included in a wrap token embedded a GSS-EAP token from the acceptor
(really, the KDC), would be suitable for constructing such an output
credential handle.

The initator app would then GSS_Store_cred(), if it cares to.  Or,
when the app uses the original GSS_Init_sec_context(), the mechanism
or the mechglue could just do it for the app according to local
policy.

Nico
--
_______________________________________________
abfab mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/abfab

Reply via email to