Abstract The World Wide Web

The World Wide Web is the largest and most successful hypermedia system developed. The popularity and scalability of the Web is due to its simplicity and its expandable architecture, enabling developers to incorporate new emerging technologies. However the simplicity renders it impossible to build collaborative applications and sophisticated structures in a pure Web environment. This has motivated many hypermedia researchers to integrate hypermedia systems with the existing Web, and thereby adding advanced hypermedia technology on top of it. However in order for such systems to succeed they must be able to scale with the Web, and up until now all approaches have been based on a client-server architecture which imposes some limitations on the scalability of the systems. This thesis investigates a new approach to the development of scalable hypermedia systems for the Web, which utilizes a peer-to-peer network architecture to overcome the limitations of the client-server approach. Since this approach is largely un-investigated, two investigation methods, simulation and prototyping, have been deployed which each serve a different purpose. The HyperPeer prototype forms a basis for the investigations regarding combining the hypermedia and peer-to-peer field. The prototype provides a plug-in framework which enables easy extension of the core hypermedia functionality and thus provides a solid base for experimentation. This thesis describes the system along with the experiences gained and problems related to development of hypermedia systems in a peer-to-peer context. A tailor-made search algorithm is introduced in order to enable search for hypermedia data in a peer-to-peer network. The algorithm is guided by resemblance and seeks to cluster resembling peers by rearranging the topology of the overlay network. Furthermore it seeks to filter the search results in order to minimize unwanted data like spam. To investigate the algorithm a peer-to-peer network simulator has been developed. The simulator provides a framework for simulating search protocols in peer-to-peer networks, and currently holds two reference protocols for comparison.