Empirical evidence from information system projects demonstrates that 'commuting loop' structures are endemic in information models. These structures correspond to the 'commuting diagrams' of category theory which are a way of expressing equations. The concept of commuting loop reflects natural constraints, the essential phenomena we believe, of information systems. To begin with, the concept takes advantage of data structure functional dependency as a natural foundation for system structure. That is, information system structure is governed by relationships and attributes. Once normalised, information system structure is essentially the structure of functions between entities and value sets. Category theory is thus applied to focus on the higher order issues that govern this function structure, rather than the entities themselves. It turns out that commuting loops provides a new higher order normalisation technique not only for the constraints over very large information systems, but for partitioning the system itself. Process structure is defined by dependencies between actions on entities. These are governed in turn by functional dependencies between entities which we now interpret as providing a causal basis for actions. Thus commuting loops capture a notion of causality within information systems. The intriguing outcome is that despite the apparent esoteric origins of commuting diagrams, their application appeals as simple, intuitive and natural from both a business and scientific point of view. What is more, the application of commuting loops in practice works. At the conceptual level, the commuting loop concept enables business process engineering to be based on the data architecture that underpins the organisation's information systems.
|Journal||IFIP Transactions A: Computer Science and Technology|
|Publication status||Published - 1994|