A Situated Approach of Roles and Participation in Open Source Software Communities

Abstract

Our research aims at understanding the various forms of participation in Open Source Software (OSS) design, seen as distributed design in online spaces of actions–discussion, implementation, and boundary between these spaces. We propose a methodology—based on situated analyses of a formal design process used in the Python project—to identify the distribution of actual roles (implementation, interactive, group, and design oriented) performed by participants into and between the spaces (defining boundary spaces). This notion of roles is grounded in collaborative design activities performed online by participants. This way, our findings complete the core-periphery model of participation in OSS. Concerning the distribution of roles between spaces, we reveal a map of participation in OSS: The majority of participants are pure discussants, but all participants in the implementation spaces do also act in the discussion space, and few participants act at boundary spaces. Concerning the distribution of roles between participants in the discussion space, we reveal that interactions are structured by a central hub (occupied by key participants) and that, whereas design-oriented roles are spread among all participants, group-oriented roles are performed by one or two participants in the respective spaces and at their boundary. Finally, combination of roles reveals five individual profiles performed by participants. Our approach could be extended to other design situations to explore relationships between forms of participation—in particular, those revealing use-oriented contributions—performance, and quality of the design product. Finally, it could be a basis for specifying tools to monitor and manage community activity for both research issues and support of online community.

NOTES

Notes

¹ A group of people connecting via the Internet with a common goal (Preece, 2000).

² Coding activity may be mostly individual, whereas it may be collective in some specific methodology, such as pair programming, for instance.

³ We previously made a distinction between three spaces named discussion, documentation, and implementation. However, we believe that it is ambiguous to refer to a single space embracing the proper implementation composed of documentation and coding.

⁴ Supported by the Concurrent Version System or by Subversion.

⁵ The PEP document is written to describe a new language feature: It is intended to provide a concise technical specification of the feature, the rationale for the feature, and a reference implementation.

⁶ The real difference in OSS and proprietary software situations may lie in more open coordination practices, that is, auto-attribution of tasks (Crowston et al., 2007).

⁷ Contrary to other studies analyzing online discussions, we have used the quotation link rather than the reply to link to reconstruct the discussions. In a previous publication (Barcellini, Détienne, Burkhardt, & Sack, 2008), we have shown that the organization of messages according to the quoting link is relevant to reconstruct the thematic consistency of online discussions and to understand the interactions between participants in terms of verbal turns.

⁸ The initiator of the discussion.

⁹ Subversion systems that trace all revisions of code, their content and authors.

¹⁰ For the record, broad traffic of these lists is as follows: python-list (5,370 participants posting 51,495 messages); python-dev 451 participants posting 8,955 messages).

¹¹ These commits and revisions are one part of the implementation space: We do not consider actions into issue tracking systems, as the Python project did not use it at the beginning of the process. Moreover, we do not consider branches in the decimal.py modules.

¹² This requirement is filled in our case, as we have 42 cells (21 participants vs. design and group-oriented activities) and four “0.”

¹³ Data represented are the same data than in Figure 4.

¹⁴ He posted around 25% of messages in both lists and performed 9% (4/44) of code revisions.

¹⁵ He is the second poster just after the champion (52/405, 13% of messages) and the third contributor in the coding space (1/44, 2% of modifications).

¹⁶ RH contributes to 77% (34/44) of all the code revisions, and he is the fourth contributor in the design space (7%, 27/405 messages).

¹⁷ Small size for the participants involvement corresponding to the first and second quartile (0–14 activities), mean size for the median quartile (15–87 activities), and larger size for the fourth quartile (equal to or more than 88 activities).

¹⁸ Only the relations greater or equal to the third quartile of the distribution of citations are represented (75% of discussions). The smallest arrows represent two mutual citations between participants, to represent superior strength we use arbitrary thickness.

¹⁹ As defined in the theoretical section, an edge is said to be a bridge if deleting it causes its endpoint to lie in different components of a network.

Background. This article is based on the Ph.D. of the first author.

Acknowledgments. We wish to thank all the members of the Python community for their kind welcome.

Funding. We thank the French Ministry for Research and INRIA for funding this research.

HCI Editorial Record. First manuscript received December 2, 2011. Revisions received September 12, 2012, March 15, 2013, and April 3, 2013. Accepted by Ruven Brooks. Final manuscript received May 27, 2013. — Editor

Additional information

Notes on contributors

Flore Barcellini

Flore Barcellini ([email protected]) is an ergonomist with an interest in cooperative and mediated activities and online communities; she is an associate professor in the Research Center on Work and Development (CRTD) of Le Cnam.

Françoise Détienne

Françoise Détienne ([email protected]) is a cognitive psychologist and ergonomist with an interest in collaborative design, technology-mediated collaboration, and online communities; she is a research professor at CNRS in the Department of Social, Economics and Human Sciences of LTCI (CNRS - Telecom ParisTech).

Jean-Marie Burkhardt

Jean-Marie Burkhardt ([email protected]) is a psychologist and ergonomist with an interest in collaborative design, mixed realities, and learning technologies; he is a research professor in the Laboratory of Driver Psychology of IFSTTAR - Institut Francais des Sciences et Technologies des Transports, de l'Aménagement et des Réseaux.