Re: [TLS] Wrapping up cached info

[Nicolas Williams <[e-mail]>]

On Thu, May 27, 2010 at 11:10:21AM -0500, Brian Smith wrote:
> Nicolas Williams wrote:
> > > How would the server indicate this to the client?
> > 
> > There's no need for the server to indicate anything to the client here.
> > 
> > If a client wouldn't attempt user authentication w/o re-negotiation then
> > it shouldn't send a hash of its user cert / cert chain in the initial
> > handshake.  If a server considers user cert TAs to be confidential
> > (highly unlikely, but let's suppose) then the server shouldn't list
> > their hashes in initial handshakes.  It's that simple.
> 
> AFAICT, that wouldn't stop a compliant (to the current draft) client from
> sending the hashes. Even if the server previously indicated that it didn't
> support the extension and/or didn't support caching certain information
> items, a complaint (to the current draft) client is still allowed to send
> those hashes:

The client can do stupid things, of course.  The point is to have text
that says "the client MUST NOT" do stupid things.

Neither the client nor the server should send all their cached hashes on
initial handshakes.  For one, that could be an enormous amount of
useless data.  Only possibly-relevant cached object hashes should be
exchanged.  The "possibly-relevant" part includes the consideration of
whether a cached object is relevant to initial vs. re-negotiation
handshakes.  I don't think we need to change the protocol to get this
right; at most we'll need to add text about this.

Nico
--