تجویزی، توضیحات، بازتاب : شکل زمینه بهبود فرایند نرم افزار

This article reviews 322 representative contributions to the software process improvement (SPI) literature. The contributions are categorised according to a simple framework: whether their primary goal is prescriptive (to tell SPI professionals what to do), descriptive (to report actual instances of SPI programs in software organisations), or reflective (theoretically analytical). The field is found to be rather dominated by one approach (the capability maturity model (CMM)) and heavily biased towards prescriptive contributions. Neither of these trends is necessarily beneficial, and it is argued that more theoretically reflective contributions could encourage a diversity of approaches which might also benefit practitioners.

Software process improvement (SPI) is an applied academic field rooted in the software engineering and information systems disciplines. It deals primarily with the professional management of software firms, and the improvement of their practice, displaying a managerial focus rather than dealing directly with the techniques that are used to write software. To date, it has been primarily practised and studied in America, Scandinavia and Australia. In terms of its theoretical heritage, SPI is equally indebted to the software engineering tradition and the Total Quality Management movement (Deming, 1982; Juran & Gryna, 1988). Classical SPI techniques (such as those built upon the capability maturity model (CMM)) relate software processes, standardisation, software metrics and process improvement. However, the field has also expanded to include other approaches (such as the software factory approach) and (at first sight unrelated) issues such as the personal discipline of software engineers and commitment. SPI stakeholders include SPI practitioners (who are responsible for improvement programs), software supplier organisations and the organisations they contract for, government bodies sponsoring research, academics and consultants. Many of the major contributions to SPI originate from the Software Engineering Institute (SEI) at Carnegie Mellon University (where Watts Humphreys has played a major innovative role). The Institute is industry-facing and supported by the American Department of Defense, whose principle interests are to identify competent software suppliers and ensure the delivery of high quality software. Analysis of the SEI's income for 2002 (the latest available year) showed that 65–73% of their income came from American Department of Defense or American military sources.1 Many consultancy, teaching and licensing activities are also associated with the SEI, and their directly-sponsored project work amounted to half their income. In this article, we develop a picture of the shape of the SPI field by analyzing it against three categories representing forms of writing. Prescriptive contributions are primarily concerned with informing SPI pratitioners how they can carry out software process improvement initiatives. Descriptive contributions are primarily concerned with describing those initiatives. Reflective contributions are primarily concerned with setting the other contributions in a theoretical context, or developing theory. The analysis framework is described more fully in Section 2.2. In principle this simple framework could be used to analyse any applied academic field. By developing such a picture, we expose some strengths and weaknesses of the field and contribute to focusing the direction of future research.

In this paper, we assessed the shape of the SPI field by categorizing a substantial number of contributions to its academic literature as prescriptive, descriptive or reflective. Though there could be many objections to the way we built the database, the analysis framework, and to the way we categorised the contributions, we think our two main conclusions would hold under any analysis of the field. We conclude that the field is rather dominated by CMM, and that it is a rather prescriptive and non-reflective field. Neither of these trends is necessarily beneficial to the field as a whole. Whilst acknowledging the very many successes and innovations of the SEI with CMM, it has never been clear that it is widely appropriate or successful outside its natural habitat (though it is very widely known), and its very success, coupled with the lack of serious reflective challenges to it, may have stifled the development of a more multi-faceted range of approaches (such as exist in the systems development field). Such a multi-faceted range of approaches could well be beneficial in the wide range of cultural and situationally different circumstances under which the software is built. In order to build on the existing reflective work in the SPI, we would encourage the following types of SPI research: Descriptive studies of SPI initiatives carried out by trained independent researchers. Statistical studies across SPI projects withsound methodological foundations carried out by independent researchers.Theoretical analysis of such descriptions using theories from related disciplines. Reflective cumulative accounts of trends. The building of theoretical accounts of improvements in software construction based on relevant theories, and/or independently researched descriptions of actual projects. Suchforms of researchmay redress the balance a little, and ensure a better balance between theory and practice in the future.