The semantic Web and Web services have emerged as a new paradigm for knowledge-based applications, both human and machine controlled. The coordination of semantic clients or services is necessary to achieve goals only possible from the combination of knowledge based activities. Triple space computing is a new coordination paradigm based on semantics and tuplespaces which could be used for the coordination of semantic Web clients and services. Triple space computing substitutes synchronous message exchange for persistent publication of knowledge in virtual data stores called tuplespaces. A tuplespace based architecture can handle the scalability, heterogeneity and dynamism which arises in highly distributed environments. In this paper, we describe the important architectural decisions in our implementation of a semantic tuplespace based middleware. We report on our practical experiences, and the benefits and challenges of triple space computing as a new paradigm for semantic Web and Web service communication.