Richard W. Selby’s research while affiliated with University of California, Irvine and other places

Active process support is being created to help human developers, including non-technical managers, coordinate their activities while integrating commercial project planning tools. Analysis and testing tools are being created to provide high assurance in software, through provision of prerun-time analysis, test case development, test result checking, and test process management tools. Powerful program analysis tools are being created which enable early detection of subtle coordination errors in concurrent systems composed of heterogeneous parts. Evolvable software architectures, first in the domain of user interface software, are being created to enable a more component-based software economy. World-Wide-Web (WWW) and hypermedia technology is being developed to foster easy information access, connections between software artifacts, and human understanding of complex systems. The UCI/Purdue team developed these technologies and validated them through interaction with the aerospace community, military organizations, the commercial software world, and the open Internet community.

Microsoft Corp. has probably tackled more PC software projects than any other company in the industry. Microsoft Corp.'s philosophy for product development has been to cultivate its roots as a highly flexible, entrepreneurial company and not to adopt too many of the structured software-engineering practices commonly promoted by such organizations as the Software Engineering Institute and the International Standards Organization. Microsoft Corp. has tried to scale up a loosely structured small-team style of product development. The objective is to get many small parallel teams or individual programmers to work together as a single relatively large team in order to build large products relatively quickly while still allowing individual programmers and teams freedom to evolve their designs and operate nearly autonomously. This article summarizes how Microsoft Corp. uses various techniques and melds them into an overall approach that balances flexibility and structure in software product development. The authors label Microsoft Corp.'s style of product development, the synch-and-stabilize approach.

OVERVIEW: Today Microsoft owns the operating systems and basic applications programs that run on 170 million computers. Beyond the genius of co-founder/CEO Bill Gates, what accounts for the company's dramatic success? From two years of on-site observation and interviewing at Microsoft headquarters, the authors identify seven complementary strategies that characterize how Microsoft competes and operates: Find smart people who know the technology and the business; organize small teams of overlapping functional specialists; pioneer and orchestrate evolving mass markets; focus creativity by evolving features and "fixing" resources; do everything in parallel, with frequent synchronizations; improve through continuous self-critiquing, feedback and sharing; attack the future! Moreover, Microsoft's "synch-and-stabilize" approach to product development enables the company not only to build an increasing variety of complex features and end-products for fast-paced markets with short life cycles, but also to shape evolving mass markets and foster organizational learning.

Software systems contain multiple types of interrelations among components — data, control, and sequencing, among others. We are developing interconnectivity analysis techniques that derive multiple views of the structure of large-scale software systems. These techniques calculate interconnections among components and then recursively group the components into sets according to their degree of interconnection. These techniques are especially suited to large-scale systems (e.g., > 100,000 lines) since numerous types of interconnections can be determined automatically in a tractable manner. Interconnectivity analysis techniques produce visualizations of system structure and can be used to document systems, model their evolution over time, compare system structure, guide regression testing, or localize error-prone structure. This article summarizes two studies using interconnectivity analysis. In earlier work, one such technique was applied effectively in a feasibility study to characterize the error-prone components in a large-scale system from a production environment. Routines with the highest coupling/cohesion ratios had 8.1 times more errors per 1,000 source lines of code that did routines with the lowest coupling /cohesion ratios. A second validation study is currently underway. In this study, we are applying interconnectivity analysis techniques to the design specification of a large distributed command and control system. Tools supporting interconnectivity analysis will be integrated into the Amadeus measurement-driven analysis and feedback system.

We have only begun to understand the experimental nature of software engineering, the role of empirical studies and measurement within software engineering, and the mechanisms needed to apply them successfully. This volume presents the proceedings of a workshop whose purpose was to gather those members of the software engineering community who support an engineering approach based upon empirical studies to provide an interchange of ideas and paradigms for research. The papers in the volume are grouped into six parts corresponding to the workshop sessions: The experimental paradigm in software engineering; Objectives and context of measurement/experimentation; Procedures and mechanisms for measurement/experimentation; Measurement-based modeling; packaging for reuse/reuse of models; and technology transfer, teaching and training. Each part opens with a keynote paper and ends with a discussion summary. The workshop served as an important event in continuing to strengthen empirical software engineering as a major subdiscipline ofsoftware engineering. The deep interactions and important accomplishments from the meeting documented in these proceedings have helped identify key issues in moving software engineering as a whole towards a true engineering discipline.

Software systems contain multiple types of interrelations among components — data, control, and sequencing, among others. We are developing interconnectivity analysis techniques that derive multiple views of the structure of large-scale software systems. These techniques calculate interconnections among components and then recursively group the components into sets according to their degree of interconnection. These techniques are especially suited to large-scale systems (e.g., > 100,000 lines) since numerous types of interconnections can be determined automatically in a tractable manner. Interconnectivity analysis techniques produce visualizations of system structure and can be used to document systems, model their evolution over time, compare system structure, guide regression testing, or localize error-prone structure. In earlier work, one such technique was applied effectively in a feasibility study to characterize the error-prone components in a large-scale system from a production environment. Tools supporting interconnectivity analysis will be integrated into the Amadeus measurement-driven analysis and feedback system.

Software measurement and experimentation provide a cross-cutting foundation for software understanding, analysis, evaluation, and improvement. Effective measurement and experimentation requires a variety of issues to be addressed ranging from goal specification to metric definition to data interpretation. This paper focuses on a subset of the measurement and experimentation issues related to frameworks, mechanisms, and infrastructure. In particular, the paper highlights research issues or results in these areas: frameworks for measurement and experimentation, existing measures, determining appropriate measures, data collection, experimental designs, and infrastructure for measurement.

The software engineering field requires major advances in order to attain the high standards of quality and productivity that are needed by the complex systems of the future. The immaturity of the field is reflected by the fact that most of its technologies have not yet been analyzed to determine their effects on quality and productivity. Moreover, when these analyses have occured the resulting guidance is not quantitative but only ethereal. One fundamental area of software engineering that is just beginning to blossom is the use of measurement techniques and empirical methods. These techniques need to be adopted by software researchers and practitioners in order to help the field respond to the demands being placed upon it. This paper outlines four paradigms for experimentation and empirical study in software engineering and describes their interrelationships: (1) Improvement paradigm (2) Goal-question-metric paradigm, (3) Experimentation framework paradigm, and (4) Classification paradigm. These paradigms are intended to catalyze the use of measurement techniques and empirical methods in software engineering.

Metric-driven classification models identify software components with user-specifiable properties, such as those likely to be fault-prone, have high development effort, or have faults in a certain class. These models are generated automatically from past metric data, and they are scalable to large systems and calibratable to different projects. These models serve as extensible integration frameworks for software metrics because they allow the addition of new metrics and integrate symbolic and numeric data from all four measurement abstractions. In our past work, we developed and evaluated techniques for generating tree-based classification models. In this paper, we investigate a technique for generating network-based classification models. The principle underlying the tree-based models is partitioning, while the principle underlying the network-based models is pattern matching. Tree-based models prune away information and can be decomposed, while network-based models retain all information and tend to be more complex. We evaluate the predictive accuracy of network-based models and compare them to the tree-based models.