Pathological Cases for DFGs

March 19, 1999
(Modified: March 19, 1999)

The purpose of this page is to warn of potential pit-falls in using DFG's, to discuss unusual cases that are in the fringe of the rule-set for DFG execution, and to solicit opinions on how we might want the SCHEDULER to handle these ill-defined situations.

At the very least, they are very interesting to consider. The extreme cases are often useful in helping to think about the paradigm and improving understanding.

Case 1: Node with no input arcs:

If a node has no input arcs, then it has nothing to wait for, and it's firing status always appears ready-to-fire. This results in an infinite loop.

Is this the interpretation and behavior we all prefer and expect?

User response indicated the following preferred behavior:

	Nodes having no input arcs should fire once at start-up,
	and never fire again.
This behavior is consistent with the START node, which has no inputs and was formerly considered to be a special case. Now, your graphs can, in effect, contain any number of separate start-nodes, and they can be called anything you want.

Status: Implemented as specified.