In 1984, Peter Patel-Schneider published a paper [1] entitled Small can be Beautiful in Knowledge Representation in which he advocated for limiting the expressive power of knowledge representation formalisms in order to guarantee good computational properties and thus make knowledge-based systems usable as part of larger systems. In this paper, I aim at showing that the same argument holds for the Semantic Web: if we want to give a chance for the Semantic Web to scale up and to be broadly used, we have to limit the expressive power of the ontologies serving as semantic marking up of resources. In addition, due to the scale of the Web and the disparity of its users, it is unavoidable to have to deal with distributed heterogeneousontologies. In this paper, I will argue that a peer-to-peer infrastructure enriched with simple distributed ontologies (e.g., taxonomies) reconciled through mappings is appropriate and scalable for supporting the future Semantic Web.