Issues in object database management

While the availability of commercial systems from several vendors indicates maturity in object database management technology, there are numerous issues which remain. This panel will attempt to expose and discuss several of these issues. Part of the performance advantage realized by object database management systems comes from linking application programs with the database management system, and the use of large virtual memory caches. This is acceptable in engineering applications where previously large amounts of data were read from the file system into an application program's data space. However, the potential impact on database integrity of giving application programs direct access to very large database management system caches will be of great concern in commercial applications. How can these concerns be addressed with minimal impact on the performance advantage of object database management? There appear to be two distinct approaches to object query languages: extensions to SQL and programming language extensions. SQL extensions might provide a fast path to a standard, but would have the traditional impedance mismatch problems associated with embedded query languages. Language extensions are elegant in that they use the same syntax as the programming language and do not suffer from impedance mismatch problems. However, language extensions would probably make standards more difficult to arrive at as it would require the coordination of extensions to multiple languages. Which of the approaches is most appropriate? In which order should these two approaches be addressed by the object database vendors and standards bodies? One might argue that a good deal of research in relational theory has had little or no impact on commercial relational systems (e.g., relational dependence theory). From a vendor's perspective, what are the hard, interesting research issues whose resolution would allow you to build better systems?