In spite of its difficulty, communication between philosophy and classification -- and more precisely between ontological research and knowledge organization --  is conceptually very intriguing, and implies a lot of problems for scientific research. One of these, which is the object of discussion of this paper, is the theme of the "something" (the matter) of classification. After a quick examination of the debate around the alternative between classification of documents and of entities, the paper shows how every kind of rigorous classification must be supported by an ontology. An example of an ontological approach to knowledge classification is the implementation of the theory of levels of reality. Nicolai Hartmann's idea of a stratification of the real world is individuated as a robust support to the aim. However a theory of levels needs a set of general categories, common for all the levels. The aim of exploring the fundamental or top categories is pursued through a combination of the critical ontology elaborated by Hartmann and some notions from General System Theory.