An Outsider’s View of Dataflow by Allan Gottlieb Ultracomputer Note #164 July, 1989 ABSTRACT An outsider’s view of the current state of dataflow research is presented. Two related observations are made. First, the field has matured from emphasizing proof of concept to pragmatic concerns needed to obtain high performance. Second, dataflow research is moving closer to mainstream parallel processing. Several examples of dataflow research supporting these observations are given. They include two-stage enabling rules and other eclectic scheduling disciplines, hardware and software throttling of parallelism, non-functional and (possibly) non-deterministic execution semantics, tradi- tional memory management, and von Neumann compatibility. ________________________________ Supported in part by Department of Energy contract DE-FG02-88ER025053.