Clustering is an important research topic for wireless sensor networks (WSNs). A large variety of approaches has been presented focusing on dierent performance metrics. Even though all of them have many practical applications, an ex- tremely limited number of software implementations is avail- able to the research community. Furthermore, these very few techniques are implemented for specic WSN systems or are integrated in complex applications. Thus it is very difficult to comparatively study their performance and almost impos- sible to reuse them in future applications under a dierent scope. In this work we study a large body of well estab- lished algorithms. We identify their main building blocks and propose a component-based architecture for developing clustering algorithms that (a) promotes exchangeability of algorithms thus enabling the fast prototyping of new ap- proaches, (b) allows cross-layer implementations to realize complex applications, (c) oers a common platform to com- paratively study the performance of dierent approaches, (d) is hardware and OS independent. We implement 5 well known algorithms and discuss how to implement 11 more. We conduct an extended simulation study to demonstrate the faithfulness of our implementations when compared to the original implementations. Our simulations are at very large scale thus also demonstrating the scalability of the original algorithms beyond their original presentations. We also conduct experiments to assess their practicality in real WSNs. We demonstrate how the implemented clustering algorithms can be combined with routing and group key es- tablishment algorithms to construct WSN applications. Our study clearly demonstrates the applicability of our approach and the benets it oers to both research & development communities.