Implementational View of Blackboard

~ ~ This section details the difference between the logical view of a blackboard structure for controlling the agents, and the implementational view of an actual set of agents written in CLIPS. Figure shows the abstract view of this system, while Figure shows a view which is more concerned with implementational issues.

In Figure , the blackboard is shown to be a combination of the CLIPS fact list and the COOL object list. These lists are available to the user through the C program and Motif, and to each of the sets of agent rules. As such, each of the agents has the access to whatever information it may find relevant. The agent rules are separated into different files and read into the CLIPS shell on program execution. So there are seemingly separate agents, reacting to the same information and providing each other with information from across the different aspects.

There are ways in which this implementation is not consistent with the blackboard model:

1. The agents act differently from the way the user acts. They react to the user, and continue to act until all agents have had a chance to react to the user's action. In this sense, the user is treated as a different kind of agent, and this is inconsistent with the ideal model. Ideally, the user would act just like any other agent, without the actions of the others revolving around the user's actions.
2. The execution of CLIPS is controlled from outside of CLIPS. The C code has to be invoked to send a ``RunCLIPS'' command to execute the rules. If the system were a blackboard it would inform the agents when a new fact appeared, and they would be able to react independent of the user's actions. CLIPS would be running constantly, and not have to be called upon to run only when new facts have been asserted.
3. Even though the agents seem to be separated, the rules are still read in and organized together and indexed by CLIPS. A real blackboard system would be responsible for indexing them and keeping them separate.