[coyotos-dev] Timeouts (was Sleep for interval)

[Jonathan S. Shapiro]

On Fri, 2007-10-05 at 09:29 -0700, Charles Landau wrote:
> At 11:56 AM -0400 10/5/07, Christopher Nelson wrote:
> >there are places where it seems impossible to get away from 
> >timeouts.  As an example, trying to connect to an unresponsive host. 
> >You will never get a "fail" message.  At some point you have to stop 
> >trying.
> 
> At some point you have to stop trying, but that point need not be 
> defined by the passage of a pre-specified amount of time. Unless 
> there is a reason for a specific timeout (the disk seek example), you 
> should pass the responsibility to stop trying back up to your client, 
> who might have better knowledge.

I agree with Charlie, but I think there may be a better way to frame
this:

  For each type of recoverable error, there will tend to emerge a
  very short list of places where recovery can be achieved. These
  are generally in some control flow scope that is well outside
  the point where the error occurs, and that is where the responsibility
  for recovery should be placed.

  *Because* the recovery criteria cannot (in general) be locally
  determined, the IPC primitive is not a particularly good place
  to embed the test that triggers the recovery.

So: yes, the error usually needs to propagate outwards up to a point,
but for recoverable errors that will generally NOT turn out to be the
user command line.
-- 
Jonathan S. Shapiro, Ph.D.
Managing Director
The EROS Group, LLC
www.coyotos.org, www.eros-os.org