The future of autonomous decentralized systems

Over the past fifty years the size of computer applications, as gauged by any reasonable measure, has increased exponentially. In contrast, our cognitive capabilities for understanding have not changed. This discrepancy has forced us to look for more abstract representations of systems to reduce the number of elements we have to consider simultaneously in order to understand the system behavior. In this contribution we argue that the next unit of abstraction, beyond the component concept, will be a subsystem that performs a specified service autonomously despite the occurrence of disturbances that were not foreseen in the original design.