Abstract Resolving the Integrity/Performance Conflict

Future applications will require integrity of complex, persistent data in the face of hardware and program failures. Thor [6], a new object-oriented database, offers a computational model that ensures data integrity in a distributed system without sacrificing expressiveness or performance. Perhaps surprisingly, compiler technology is important to making this work well. 1 Traditional Approaches Currently, writing applications that share complex, persistent data across a network is too hard, because the underlying software substrate is not available. No existing system combines high performance, complete data integrity, and a sufficiently powerful data model. Allowing multiple applications to share complex persistent data creates problems because of the desire for protection. Increasingly, persistent data contains semantic information. It may only be accessed in particular ways that ensure its integrity is not violated. Data may be accessed by multiple users, or shared across a network which can experience failures. Failures of hardware, or of users to properly access objects, cannot be allowed to compromise persistent data integrity. The conventional approaches to providing these services have been operating systems, databases, and persistent programming languages. Each o