Re: [Fresco-devel] Pinging stopped clients

Neil Pilgrim wrote:
> Please see:
> - task77 Make calls into client code non-blocking
> [issues.fresco.org/task77]
> - task78 Redesign server connection-tracker [issues.fresco.org/task78]
> 
> Comments and updates gratefully accepted :)

Don't ping clients. Remove the ping method.

To Fresco a Client does nothing but access the Kits to produce the 
Graphics in the scene graph. Hence:

i) A client could connect to contruct a scene graph displaying data, 
then exit once the SG is built.
ii) Any Graphic in the scene graph does not need to be associated with 
any ClientContext.

When we want to "ping" a client, what we really mean is that we want to 
verify the integrity of the scene graph in two ways:
i) the Graphics can be expected to draw in a timely fashion
ii) accessing Graphics and Commands will not time out

Pinging the client assures neither. It's not even a guarantee that a 
given Command hasn't disappeared. (Commands being an obvious example of 
a client-side object that we need to access, from any Trigger.) The ping 
method should be removed.

The question is, how do we verify the integrity of an entire distributed 
system? Even if we were to traverse our scene graph (without drawing 
it), just to make certain that it i) throws no exceptions; ii) is 
acyclic, we still haven't shown that this will be the case one 
millisecond from now.

Nick Lewycky