In this paper we analyze the problem of schema matching, explain why it is such a "tough" problem and suggest directions for handling it effectively. In particular, we present the monotonicity principle and see how it leads to the use of top-K mappings rather than a single mapping.