This proposal describes a system architecture that we propose to implement as a programmer's toolkit for the triggered distribution of information within a distributed object-oriented database. The system architecture permits sudden appearance and disappearance of data objects, understands that the communication infrastructure can be temporarily partitioned, and guarantees that rule firings cannot cause unstable, non-terminating sequences of updates. The architecture permits a spectrum of data consistency algorithms and database transaction models. The data object structure allows procedures within the object to be preceded by a pre-condition that can check permissions, legitimate input values and data integrity prior to execution. Post-conditions implement consistency checking, replication of objects and processes, and triggers that further activate other procedures or other objects. Triggers enable versions of objects. The toolkit consists of routines to communicate updates and register rules and triggers between databases. The architecture supports the efficiency discovery of objects that merit attention. It permits dynamic changes in the rule set, yet provides an efficient indexing mechanism to avoid evaluating triggers needlessly.
