This paper investigates several approaches for qualitative spatial knowledge representation on the Semantic Web, by using RCC8 relations. We discuss several issues arising when representing RCC8 in OWL DL, e.g., the lack of required features like role reflexivity, role Boolean operators, and role inclusion axioms. We conclude that, although some of these features are to be included in the new version of the OWL standard, OWL 2, this language still lacks the expressive power to support role negations, conjunctions, and disjunctions, and complex role inclusion axioms.