TITLE: ``An Architecture (and Implementation) of a Cooperative Database System''

ABSTRACT:

Current database systems (and information systems in general) are uncooperative. They respond to a query with just the query's evaluation, the answer set. This can sometimes be misleading to the user.

For database systems to be more cooperative, they sometimes must respond with more information than just the answer set to the user's query. We call such responses cooperative, and such database systems cooperative.

We have been designing and implementing such a cooperative database system (CDBS). We call our prototype CARMIN. The system is a shell that extends a relational database system (RDBMS). By building on top of an RDBMS, we reuse existing relational technology (we do not have to reinvent it), we inherit the optimizations that RDBMSs make, and our system can be immediately used with existing relational databases.

In this talk, I introduce the core cooperative techniques that we support. I then discuss an architecture for a CDBS that interfaces with an RDBMS. The architecture must accommodate the computational complexities of the various cooperative techniques appropriately, and it must handle the constraints introduced by the RDBMS interface. I will also indicate the technical challenges that arise in an implementation, such as algorithms that are needed to actuate certain cooperative techniques efficiently. (However, in this talk I will not discuss the implementation issues in any detail, although we address them in the CARMIN project.) The architecture also introduces an appropriate modularity for a CDBS which allows the system to be constructed incrementally.

Everyone is welcome to attend.