This paper defines a new, context-driven programming model for pervasive spaces. Existing models are prone to conflict, as it is hard to predict the outcome of interleaved actions from different services, or even to detect that a particular device is receiving conflicting instructions. Nor is there an easy way to identify unsafe contexts and the emergency remedy actions, or for programmers and users to grasp the complete status of the space. The programming model proposed here resolves these problems by improving coordination by explicitly defining the behaviors via context, and providing enhanced safety guarantees as well as a real-time, at-a-glance snapshot of the space’s status. We present this model by first revisiting the definitions of the three basic entities (sensors, actuators and users) and then deriving at the definition of the operational semantics of a pervasive space and its context. A scenario is provided to demonstrate both how programmers use this model as well as the advantages of the model over other approaches.