Beyond Data Dictionaries: Towards a Reflective Architecture of Intelligent Database Systems

The main conjecture of this contribution is that forthcoming intelligent database systems — in particular future DOOD systems — should be designed in such a way that a major part of the services they provide are implemented using these same services in a bootstrapping-like manner. We call such an approach “reflective”, as is often done by researchers in AI and programming languages. Data dictionaries, being part of any reasonable database system today, exhibit the reflective principle in a nutshell, if they are implemented by means of the same data structures that hold application data. However, even for data dictionaries the reflective implementation is often abandoned for performance reasons. Applying reflection for more advanced and ambitious purposes, up to integrity control or query optimization, is viewed even more skeptically by many, despite the conceptual elegance of the approach. On the other hand, there are a few successful approaches around today that can be interpreted as exhibiting a reflective nature. It is the purpose of this paper to identify such examples and to encourage research to invest more in the reflective style and to look for new solutions to the obstacles still ahead.