Best-match-pair is proposed as a new strategy for the control mechanism of blackboard systems. A generalised blackboard shell with the strategy as the control loop is proposed, which evaluates both solution areas and knowledge sources explicitly. This approach improves the system performance by providing a balanced status feedback between potential knowledge sources and promising solution areas for system control and replanning. It also enhances the system performance by supporting viewpoint reasoning, shadow reasoning, and various metalevel applications, e.g., viewpoint explanation, knowledge acquisition, etc.