For some time there has been an effort to bring together the scattered fragments of spatial information theory into a single, organised body of knowledge and theory. The state of this effort is first surveyed and many of the gaps and problems identified. Following this, a list of desired characteristics for a full theory is presented. Then, a new framework or paradigm which meets most of the desired characteristics is proposed. This new paradigm is coherent with previous work in bringing together spatial information theory, particular the efforts which have been recently labelled Naive Geography, but the proposed paradigm goes beyond the characteristics heretofore proposed. Some of the consequences of adopting the new paradigm are then discussed in the paper. The new paradigm integrates cognitive and geometric approaches to space, provides a framework for understanding the relationship between different data structures such as raster and vector, and situates much existing work with GIS into a well-defined sub-discipline of what might potentially be a much larger discipline.