| Distributed, dataflow-graph-based, iconic visual programming that relies on a demand-driven, eductive evaluation scheme is investigated and a working prototype environment that runs on networks of Macintosh;vLucid provides a framework for enhanced Lucid program and data visualization, and greatly simplifies the programming task by employing easy-to-learn direct-manipulation movements. It extends classical Lucid's I/O capabilities to provide for greater interactivity and operating system level interfacing and accomplishes this in a uniform and symmetric manner. vLucid also provides support for the different modes of distributed programming investigated.;Three models of distributed visual dataflow programming are discussed: the Blind Producer-Consumer (BPC) model, the Program Server (PS) model, and the Coprogramming model. The BPC model makes use of a crude explicit distributed programming technique and serves to introduce the subject. The PS model allows implicit distributed programming through both user-directed and limited automatic program component-to-processor mappings. The Coprogramming model extends this foundation by providing a basis for the sharing of both processing capabilities and program construction modules across networks.;Bidirectionality is also explored in this research as a method of reducing operator set size and interconnection densities. Two modes of particular interest are presented in detail: Static Asymmetric Bidirectionality and Dynamic Symmetric Bidirectionality. The former mode is of note because it permits data-driven operation within demand-driven programs. The latter mode relies on a relational programming foundation in which only functional modings are permitted. Operator schemas for Lucid operators with reversed, functional directionality are presented. Also, constraint programming within Lucid and employing bidirectionality is explored, and a new iterative technique for solving certain types of self-referential dataflow programs is presented. |
