The merging of grid and pervasive computing raises new challenges. Its main impact is a critical increase of the dynamicity, heterogeneity and unpredictability of users, usages and even network topology. Moreover emergency scenarii are one of the most important target applications of pervasive grids: With strong time constraints the performances become a critical point and the decisions about the distribution become a key of the efficiency. However existing grid middlewares have been designed for quite stable, homogeneous, and predictable architectures. Our proposal is a distribution decision-making support designed to be easy, usable, and profitable. We present a use case of pervasive grid and demonstrate why existing solutions are not adapted. We show how our method has been embedded in a grid web service, namely the Network Distance Service (NDS) and illustrate its benefit on the main problems of distribution: replication, placement and selection. We show experimentation results and discuss how NDS handles the new constraints involved by the merging of grid and pervasive computing.