Principled Research in Database Systems Jennifer Widom, Stanford University A surprising fraction of database research is purely systems-driven, with little regard for well thought-out, clean foundations. We argue that developing new types of database systems is best achieved by laying solid foundations first, before any systems-building begins. We look specifically at the challenges of developing new data models and query languages that are simultaneously well-defined, understandable, sufficiently expressive, similar to existing models & languages, and implementable. Detailed examples are drawn from our own work in active databases, semistructured data, data streams, and uncertain data.