Dataflow is presented as an alternative to the von Neumann model as the basis for computer system design. The need for a new semantic basis is supported by current research both in software methodologies and in computer architecture. Dataflow systems emphasize asynchronous and functional computation. We present a high-level dataflow language Id, and its companion base language. Id supports programming with streams, programmer-defined data types, and facilities for nondeterministic programming. The base language when interpreted by the unfolding interpreter generates a potentially large number of independent activites which can be executed concurrently by a dataflow machine. The unfolding interpreter seems very promising for implementation on a machine composed of large numbers of LSI processors.

Current solid state technology suggests that future computers will be highly asynchronous machines comprising small intercommunicating processors, each processor contributing its effort to some part of the ongoing computation. The functional basis of such a machine demands a totally different foundation than that of current machines and languages. Data flow has been suggested as an alternative approach to vonNe mann type machines and associated sequential languages. A major cricism of data flow in the past has been the lack of a suitable higher-level programming language for coding programs. We feel that the sketch of a higher-level language presented here not only answers the criticism but also brings an asynchronous control structure into programming languages, as well as a strong theoretical basis for such properties as modularity and verifiability of programs. The paper contains a description of the Irvine Data Flow language ID. Three complete examples of programs in ID are included in section 3. For sake of reference, a BNF grammar for ID has been included in Appendix A.