Modeling geospatial databases for GIS applications has always posed
several challenges for system analysts, developers and their clients. Numerous
improvements to modeling formalisms have been proposed by the research
community over the last 15 years, most remaining within academia. This paper
presents generic extensions (called Plug-Ins for Visual Languages or PVL) to
facilitate spatial and temporal modeling of databases. For the first time, we explain
its intrinsic relationship with an extended repository and how it has been
influenced by pragmatic lessons learned from real life projects. We describe
how we use PVLs with UML and how 16 years of fundamental research, diverse
experimentations and feedbacks from users over the world shaped our approach.
The final section presents Perceptory, a free repository-based UML+
PVL CASE developed to improve geospatial database modeling.