A method of denotational formalization of data bases, of data base management systems, and of related structures is proposed, aiming to improve their understanding, specification and rigorous investigation. The method provides a uniform treatment of different information layers: from instantaneous data bases (the first layer), via schemata and integrity rules, to classes of data base models. It unifies in one mathematical notion the apparently different notions of the semantics of stored data, semantics of data base processes, conceptual semantics of data bases, integrity semantics of data bases and denotational semantics of languages. The unification is based on hierarchies of domains of continuous mappings between different representations of information (from "less semantic" representations into "more semantic" ones).