In this paper we consider the provisioning problem of Web services. Our framework is based on the existence of an abstract composition, i.e., the way some services of different types can be combined together in order to achieve a given task. Our approach consists in instantiating this abstract representation of a composite Web service by selecting the most appropriate concrete Web services. This instantiation is based on constraint programming techniques which allows us to match the Web services according to a given request. Our proposal performs this instantiation in a distributed manner, i.e., the solvers for each service type are solving some constraints at one level, and they are forwarding the rest of the request (modified by the local solution) to the next services. When a service cannot provision part of the composition, a distributed backtrack mechanism enables to change previous solutions (i.e., provisions). A major interest of our approach is to preserve privacy: solutions are not sent to the whole composition, services know only the services to which they are connected, and parts of the request that are already solved are removed from the next requests. We introduce a specific data structure, namely Message Treatment Structure, for modeling the problem. We show the interest of this data structure to express the general principles of our framework and the related algorithms.