The development of Web standards and technologies has brought new opportunities for large-scale integration of Web content: data sources (such as XML, HTML, or PDF files), distributed applications (accessed via Web Services), and semantic information. At the same time, peerto- peer (P2P) platforms are being developed, and increasingly used for data management at the network scale. We present KADOP, a model, architecture, and system for sharing content in P2P. By building on distributed hash tables technology, XML indexing and query optimization techniques, and on the paradigm of active documents, KADOP enables the publication, discovery, and efficient exploitation of Web content in P2P. Thanks to intensional information, KADOP (i) is capable of run-time dynamic Web service composition, and (i) is open to external resources such as databases or other P2P networks.