8 A Multitasking Application

8.3 Processes

• Each node in the decision tree is actually a separate process so that parallel decisions can be pursued. You can exploit the independence of processes in the decision tree when the alternative paths involve entirely different kinds of processing or when they involve different devices. You can also have a goal pattern that matches several rules, each of which calls on a different piece of system software. Other advantages of using multiple processes for decision nodes include the ability to activate, deactivate, kill, and otherwise control each node in the tree individually.

• A supervisory process controls the growth of the decision tree by activating and deactivating the processes in that tree. This process is called the backward chainer scheduler.

• The processes in the decision tree communicate by sending messages to each other; that is, one process executes a function that causes another process to respond in some way. For example, a subprocess can send a message to its parent to indicate that a chain of reasoning has been completed.

Generated with Harlequin WebMaker