Theoretical and Practical Complexity of UML

Systems development methods have become more complex, concurrently with today’s systems. UML (Unified Modeling Language) has been criticized for its complexity, for those learning and using it. Using Rossi and Brinkkemper’s (1996) metrics, Siau and Cao, (2001) completed a complexity analysis of UML and other modeling techniques, finding that UML is more complex than other techniques. Siau, Erickson and Lee (2002) argued that Rossi and Brinkkemper’s metrics present the theoretical maximum, as opposed to a practical complexity, which must be less than the maximum. Therefore, Siau and Cao’s UML complexity analysis represents the theoretical complexity of UML. The current research proposes that a subset of UML (a kernel) composed of the most commonly used constructs, would more closely represent the complexity that practitioners face when using the language. A Delphi study is conducted using practitioners as experts, in an attempt to identify a use-based UML kernel and UML’s practical complexity.