IntroductionContinuous Design in F/OSSContinuous Design and Community KnowledgeConstruction of Community KnowledgeCommunity knowledge is constructed from a multiplicity of viewpoints, representations, and experiencesCommunity knowledge development is enacted through knowledge management environmentsThe artifact (and knowledge about it) is constructed through a dynamic network of social processesRepresentation of Community KnowledgeDocuments record viewpoints, and these viewpoints are subject to multiple interpretationsThe scope of documentation is unknowableThe fidelity of relationship between documents and experiences is uncertainThe relevant life-world is unknowableConclusionsAcknowledgementsReferencesUnderstanding Continuous Design in F/OSS Projects Les Gasser1,2 [email protected] Gabriel Ripoche1, 3 [email protected] Walt Scacchi2 [email protected] Bryan Penne1 [email protected] 1 Graduate School of Library and Information Science University of Illinois at Urbana-Champaign 501 E. Daniel St., Champaign, IL 61820, USA Phone: +1 217 265 5021 / Fax: +1 217 244 3302 2 Institute for Software Research University of California at Irvine ICS2 242, UCI, Irvine, CA, 92697-3425 Phone: +1 949 824 4130 / Fax: +1 949 824 1715 3 LIMSI-CNRS / Université Paris XI BP 133, 91403 Orsay Cedex, France Phone: +33 1 69 85 81 01 / Fax: +33 1 69 85 80 88 Abstract Open Source Software (OSS) is in regular widespread use supporting critical applications and infrastructure, including the Internet and World Wide Web themselves. The communities of OSS users and developers are often interwoven. The deep engagement of users and developers, coupled with the openness of systems lead to community-based system design and re-design activities that are continuous. Continuous redesign is facilitated by communication and knowledge-sharing infrastructures such as persistent chat rooms, newsgroups, issue-reporting/tracking repositories, sharable design representations and many kinds of "software informalisms." These tools are arenas for managing the extensive, varied, multimedia community knowledge that forms the foundation and the substance of system requirements. Active community-based design processes and knowledge repositories create new ways of learning about, representing, and defining systems that challenge current models of representation and design. This paper presents several aspects of our research into continuous, open, community-based design practices. We discuss several new insights into how communities represent knowledge and capture requirements that derive from our qualitative empirical studies of large (ca. 2GB+) repositories of problem-report data, primarily from the Mozilla project. Keywords Continuous design; Open source software; Knowledge management; Knowledge representation; Community knowledge; Specification. 1/61 Introduction In current research we are studying software maintenance work, bug reporting, and repair in Free/Open Source Software (F/OSS) communities, giving special attention to factors such as knowledge exchange, effectiveness of tool support, social network structures, and organization of project activity. More specifically, we are examining how these factors affect such outcome variables as the time taken to resolve reported problems, the ability to detect interdependent and duplicate problems, the effective scope of repairs (the degree to which the community can address the entire range of software problems that appear, rather than just selected subsets), and the general quality of software. Answering these questions is important to a number of scientific, government, or industrial communities. U.S. science agencies continue to make substantial (multi-million dollar per year) investments in complex software systems supporting research in natural and physical sciences (e.g., Bioinformatics, National Virtual Observatory) via computational science test-beds (e.g., Tera-Grid, CyberInfrastructure), which software is intended as free/open source [13,10]. E-Government initiatives around the world are increasingly advocating or mandating the use of F/OSS in their system design and development efforts [5,8]. Also, industrial firms in the U.S. that develop complex software systems for internal applications or external products are under economic pressure to improve their software productivity and quality, while reducing their costs. F/OSS is increasingly cited as one of the most promising and most widely demonstrated approach to the reuse of software [4] in ways that can make the design and development of complex software faster, better, and cheaper [22]. However, in each of these cases, there is no existing framework or guideline for how to design, manage, or evaluate F/OSS software systems and processes. Moreover we don’t have a clear picture of how these “consumers” might best expend their scarce resources for continuous system design and redesign [16], or how they might redesign their own processes to assess and take advantage of F/OSS technologies [15]. Our research process is based on qualitative and computational analysis of large corpuses of longitudinal data from Web sites, public repositories, online discussion forums, and online artifacts from F/OSS development projects. Projects analyzed include networked computer games, Internet/Web infrastructure, X-ray astronomy/deep space imaging, and academic software research. We have access to a number of large collections of current data from these projects, comprising at least six repositories of problem report/analysis activity, with a total of over 500,000 reports, containing approximately 5 million individual comments, and involving over 60,000 reporters and developers. We are using both human-based grounded-theory and computational methods (such as automated concept extraction, data and text-mining, process modeling) to empirically discover, identify and comparatively analyze patterns of continuous software design in the projects under study [14]. More specifically, we are looking for design episodes, basic design processes (such as collective sense making, negotiation, information seeking, and conflict management), design-management activities, and design-support/design-management infrastructure. For example, we are using statistical text-analysis tools to automatically extract episodes of design activity when traces of those activities can be characterized with specific “language models”. We are