CSR: SHF: Small: Propagator-Based Computing, A Programming Foundation for Decentralized Systems

Massachusetts Institute Of Technology, Cambridge MA

Investigators

Abstract

This project develops the Propagator Model, a concurrent decentralized framework designed to support computing in large, open, dynamic environments. It provides powerful tools for organizing computations to operate effectively in a world of rapidly changing and globally inconsistent data by adopting a fundamental shift in viewpoint: the things manipulated by basic computing elements are not fixed values. Rather, they are information about values, and this information is continually refined as new information becomes available. This project creates the architectural and linguistic foundations for systems that can operate effectively in environments where there is no central management, and where one cannot rely on resources being consistently available or consistently operating, and where the data is rapidly changing and globally inconsistent. Using three mechanisms implicit in the computational substrate: (1) constraint propagation, (2) partial information structures, and (3) dependencies, systems built on the propagator model automatically have the ability to support their conclusions with arguments and report on the provenance of the ingredients. They can automatically discover and use consistent subsystems of inconsistent data. This project erects a naturally concurrent and distributed model and infrastructure for computation that makes it easier to build systems that are reliable in the face of natural failures and deliberate attacks. It provides support for auditable and accountable systems that are robust and adaptable to novel applications.