On Nov 14, 2012, at 5:44 PM, Sam Hartman wrote:

>>>>>> "Alan" == Alan DeKok <[email protected]> writes:
> 
>    Alan>   For PCP, it's reasonable that the application knows the EAP
>    Alan> session has timed out.  It can re-establish it before sending
>    Alan> the next message.  Or, if messages are rare, it can wait and
>    Alan> re-establish it later.
> 
> Well, it's a bit tricky to re-establish the session if the next message
> is coming from the server but only the client can start the
> authentication.
> I think you have two general options:
> 
> 1) Decide it's OK not to re-establish the session until the client wants
> to do so
> 

This does not make sense. If we are talking about securing a protocol, that 
better be a complete security, not a partial security that's uncontrollably on 
and off. In this case, when the server wants to send a message, if it's lucky 
it'll find a live security association and send the message securely. If it's 
unlucky (there is no live SA), then bad luck -- packet is dropped w/o 
transmission. Because sending an unsecure packet does not make sense, it's at 
best useless, at worst harmful if we expect the clients to act on them.


> 2) Change things and let the server start an EAP session.
> 
> 3) Require the client to start an EAP session whenever the old session
> goes away.
> 

2&3 make sense.

I'd say the protocol shall support the capabilities needed for both 2 and 3. 
And leave it to the deployment to decide when to use them.

> The details as they apply to PCP are best discussed on that list.


Alper

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

Reply via email to