Re: [TLS] Why there should not be a TLS 2.0

On Sun, Jun 8, 2014 at 8:49 AM, Viktor Dukhovni
<viktor1dane@dukhovni.org> wrote:
> On Sun, Jun 08, 2014 at 02:27:58PM +0200, Kurt Roeckx wrote:
>
>> I would imagine that the client would like to do a "please give me
>> a tunnel to that hostname", and then get back some descriptor that
>> it can use to talk to it.  Please note that I say hostname,
>> because I think that you can have several names each which it's
>> own certificate on the same IP address.
>
> Yes, but stronger yet, one needs (ala GSSAPI + Kerberos, which for
> all its limitations gets some things right) to be able to ask for
> service@hostname, not just hostname.  The underlying Kerberos
> stack works with service/instance@REALM, which also makes sense.

See also Ethos, where the network protocol authenticates services and
users to each other. The kicker is services no longer need to
authenticate users.
>
> We could define new DANE record formats for scalable public-key
> cross realm authentication with Kerberos or a Kerberos-like system,
> and DNSSEC for a suitably secure hostname->realm mapping.

Or one could use a certificate format that supported this natively.
The fact that TLS is (mostly) tied to X509 is a real shame when it
comes to extending.

Yes, you could define extensions for certificate type and use them: at
some point SPKI may have been used this way. But as Viktor points out
GSSAPI dominates this space (unfortunately: anonymous users are hard
to rope in here).

Sincerely,
Watson Ladd

>
> --
>         Viktor.
>
> _______________________________________________
> TLS mailing list
> TLS@ietf.org
> https://www.ietf.org/mailman/listinfo/tls

-- 
"Those who would give up Essential Liberty to purchase a little
Temporary Safety deserve neither  Liberty nor Safety."
-- Benjamin Franklin