In Service-Oritented Computing, business processes are usually expressed in terms of workflows that describe control and data dependencies between loosely coupled components (usually external Web services accessible through a network). Besides sequences and parallel flows, workflows may include branches, loops, and other complex control constructs.

Knowing properties of data items and activities that access data can help with a number of design- and run-time activities in the service world. These include: workflow fragmentation (e.g. for distributed execution), data compliance checking (e.g. making sure that the given information is sufficient for a task) and robust top-down design (e.g. deriving properties for sub-workflows). In this case, we consider a set of abstract Boolean properties (i.e. attributes) whose semantics depends on the application domain. We’ll use the example of attributes that describe information content, and workflow fragmentation based on information flow across organizational domains.

The automated attribute inference relies on deriving a logic program (Horn clause) based representation of a workflow, and plugging into it the appropriate initial setups for arguments that reflect the attribues of workflow input data. Then, we apply sharing analysis (an instance of abstract interpretation), and reinterpret the results as a concept lattice to assign input attributes to intermediate data items and activities in the workflow.