Software Engineering in the Twenty-First Century

FALL 1992 71 By the year 2000, there will be a large potential market and a fertile environment for knowledge-based software engineering (KBSE). In the coming decade, hardware improvements will stimulate demand for large and sophisticated application software, and standardization of software interfaces and operating systems will intensify competition among software developers. In this environment, developers who can rapidly create robust and error-free software will flourish. Developers who deliver buggy software years behind schedule, as is typical today, will perish. To meet this challenge, software developers will seek tools and methods to automate software design. Computer-aided software engineering (CASE) is undergoing tremendous commercial growth. However, the current generation of CASE tools is limited by shallow representations and shallow reasoning methods. CASE tools will either evolve into, or be replaced by, tools with deeper representations and more sophisticated reasoning methods. The enabling technology will come from AI, formal methods, programming language theory, and other areas of computer science. This technology will enable much of the knowledge now lost in the software development process to be captured in machineencoded form and automated. KBSE will revolutionize software design, just as computeraided design has revolutionized hardware design, and desktop publishing has revolutionized publication design. This article draws on the chapters in Automating Software Design and other sources to present one vision of the future evolution of KBSE. After an executive summary and a brief history of software engineering, the role of AI technology is examined in mainstream software engineering today. Currently, AI programming environments facilitate rapid prototyping but do not produce efficient, production-quality code. KBSE technology combines the advantages of rapid prototyping and efficient code in a new programming paradigm: transformational programming, described in the subsequent part of the conclusion. In transformational programming, prototyping, validation, and modifications are done at the specification level; automatic program synthesis then translates specifications into efficient code. The following part compares the trade-offs in various approaches to program synthesis. Then, several nearterm commercial applications of KBSE technology are predicted for the next decade. To scale up from these near-term applications to revolutionizing the entire software life Software Engineering in the Twenty-First Century