Re: [TLS] Single round trip abbreviated handshake

[Adam Langley <[e-mail]>]

On Wed, Feb 3, 2010 at 7:24 AM, Ravi Ganesan <[e-mail]> wrote:
> Question: Restricting analysis to abbreviated handshake, would it not
> be the case that the session_ids and corresponding master secrets for
> exisiting session would be separate in the two clusters? i.e. unless
> you allow me to do full handshake in cluster 1 and resume it in
> cluster 2. If you do not then does this not prevent replay from one
> cluster to the other?

Although you can probably probe different Google frontend clusters to
find this stuff out, I'm going to shy away from talking about the
specifics of Google's serving infrastructure.

Keeping things general:

Session IDs need some kind of server-side storage and SessionTickets
contain arbitary server controlled data, so both could contain an
epoch number that the server uses internally when validating an nonce
for uniqueness.

So, if these schemes were to be coupled (resumption and client
supplied nonces), then the need for the client to handle the epoch
could be removed.


AGL