Computational science studies often rely on the availability of large datasets through the Web. Web services (WS) provide a convenient way for making those datasets available, since they rely on a standard and widely available technology. However, there are many ways to devise a Web service interface for a given dataset, and the resulting interfaces vary in their properties related to cohesion, distribution, flexibility and communication overhead, among other parameters. This paper explores these attributes and provides some directions on how they can be measured. Concretely, a well-...