Separating process model enactment from process execution in Provence

The conceptual distinction between the enactment of a software process model and the actual execution (or performance) of a process has been pointed out by Fernstrom (1983). Separating the two domains is essential for process support systems that aim to provide process-centered assistance without changing the project's working environment, by which I mean the set of tools and the file system utilities used by project personnel, the way the project's information is organized, etc. To achieve the separation, process support systems must monitor the actual execution of a project's processes, which takes place outside of the system, and map it to the enactment of the project's process model. Assistance can then be provided based on the process model enactment. In this manner, project personnel continue to use the working environment with which they are familiar and still benefit from the assistance provided by the process support system. Provence is an architecture for modeling, enacting, monitoring and visualizing processes that maintains the separation between process model enactment and process execution. In implementing Provence, we had to resolve three fundamental difficulties: (1) how to map the states in the process model enactment onto the states in the actual process execution, (2) how to synchronize state changes in the two domains, and (3) how to monitor the process execution while maintaining the privacy of data belonging to project personnel. This paper describes how the three difficulties were overcome in Provence. It gives some of the details of the mapping between the enactment world and the execution word, it presents the protocol used to synchronise state transitions in the two worlds, and it describes briefly how the low-level operating system events that comprise the process execution are monitored.