Web services (WS) can be seen as a pillar block for achieving electronic B2B transactions. More and more companies are using WS to achieve transactions with their partners and/or offer on-line services. These companies have to make their WS available for consultation through their own private WS registries. As a result, the number of WS registries that are made available for use can be as many as the large number of companies. This raises an old, search engine, problem in a new form: discovery mechanisms of WS are not efficient both in response times and quality of results. In this context, if appropriate solutions are not considered, "traditional" WS discovery mechanisms that consist of scanning all the registries would for instance slow down the increase rate of WS. In this thesis, we are interested in WS discovery in a distributed registry environment. To do this, we first propose to organize a registries network into communities based on functionalities offered by the WS of a registry. Then, to handle the dynamic nature of communities and their members, we define the needed management mechanisms to monitor changes and reconcile potential conflicts. Finally, we propose an approach for WS registries discovery that uses two filters to limit the search space. We first use the functional requirement of a service requester to select the appropriate registry community. Then we use the requester's non-functional requirements and his behavior to select the registries. The proposed solutions were tested by setting up a P2P network of registry communities, implementing a community manager and a discovery system.
