request-key vs pipefs

[Benjamin Coddington]

Kevin Coffman wrote:
> I would assume there is a good chance that if the context has expired,
> it is because the service ticket expired.  The service ticket is
> usually (not always) limited to the TGT's lifetime.  So if we need a
> new context, we'll probably need a new or refreshed TGT.
> 
> K.C.

Indeed.. and in that case, we're done (and we probably want to be done). 
  How is the scenario currently handled?  Is there a window in which we 
don't schedule deferred writes if the context is due to expire?

I suppose my global question is this:  should I be trying to use keyring 
support, which will result in a major branch away from what is currently 
being developed..

Are deferred writes a big enough problem -- doesn't gssd also have this 
problem with expired tickets?

Is it insane to send ticket information back into the kernel for cases 
where contexts expire before tickets?