This article presents the premises of our work on service selection in service-oriented architectures and pervasive environment. Classical registries in distributed environments use a key-value description of services, and the selection is reduced to a filtering on the key values of registered services. But in pervasive environments, the number of services increases and the user gets drown under available choices. The objective of our work is to enhance the appropriateness of service selection, by using various information on the services and the requesters. We have selected three sources of information: semantic descriptions of services, user context and preferences (own preferences and social network preferences). The storage and processing of such information is distributed among the hosts (registry and clients) depending on privacy constraints and hosts capabilities.