An approach for evolutionary database design is presented which tries to remedy some of the shortcomings of previous design methods. The approach distinguishes clearly between a conceptual and a logical database design. A conceptual schema models the relevant aspects of reality. A logical schema describes the structure of the database as generic tables, and it reflects the design decisions taken to map the objects of the conceptual schema into the generic tables. To support this strategy with tools, it is necessary to have a version concept and a mechanism for recording design decisions called protocolling. These concepts are realized in Presto, a development environment for an evolutionary design of database applications which is described. >