The first step of processing OpenStreetMap data is loading it into a PostGIS database, only to retrieve all of it later on. This talk proposes strategies and tools to generate maps without using a database, which has performance benefits as well as better control about the result.