Data mediation is a fundamental issue in information systems (IS) within which data sources are distributed and heterogeneous. Classic technics for federating databases represent a satisfiable solution in the case of loosely dynamic and autonomous databases. These properties are emphasized in grid systems.
We provide a negotiation process for an adaptative mediation mechanism. We then define a class of mediators capable of negotiating resource profiles. These latter are built from and maintained within metadata bases that are linked to each mediator. We apply our approach to the conception of a grid-based IS built upon SOA and grid service standards like WSRF et WS-DAI.