[websec] HSTS: pinning certs, other changes to TLS server authentication

[Matt McCutchen <[e-mail]>]

The following observations were made on EFF's SSL Observatory list
(https://mail1.eff.org/pipermail/observatory/2011-March/000086.html)

On Wed, 2011-03-23 at 12:56 -0400, Matt McCutchen wrote:
> HTTP Strict Transport Security does not pin the cert (it only
> prevents the user from accepting bad certs), so it is exposed to CA
> compromises.  SSH is not.

On Wed, 2011-03-23 at 12:00 -0600, Hodges, Jeff wrote:
> Yes, as presently specified and implemented in it's _draft_ form.
> 
> This could change. [...]
>
> Also, HSTS is arguably an intermediate-term (and specific-to-http)
> approach to the more general issue of network application
> advertisement of security policy. I.e. future work may supplant it.

On Wed, 2011-03-23 at 14:08 -0400, Matt McCutchen wrote: 
> Understood.  But I wouldn't propose to change that.  HSTS is the wrong
> place to fundamentally change the TLS server authentication model.

On Wed, 2011-03-23 at 12:24 -0600, Hodges, Jeff wrote: 
> I nominally disagree with the latter for various reasons, but we
> should discuss on websec@

Jeff, if you had something more to say, I invite you to say it.

-- 
Matt