Virtual organizational learning in open source software development projects
Yoris A. Au
*
, Darrell Carpenter, Xiaogang Chen, Jan G. Clark
Department of Information Systems and Technology Management, College of Business, University of Texas at San Antonio, One UTSA Circle, San Antonio, TX 78249, USA
1. Introduction
Many people can work together on a task regardless of time,
geographic location, or organizational affiliation by adopting a
virtual approach [14]. Open source software (OSS) development
projects exhibit many of the characteristics that make virtual
organizations successful, including self-governance, a powerful
set of mutually reinforcing motivations, effective work struc-
tures and processes, and technology for communication and
coordination. Examples of thriving OSS projects include Linux,
Apache, and Mozilla. Although seemingly disorganized, and
lacking monetary incentives, the development approach is
characterized by design simplicity, team work, a visible product,
and communication.
But what makes OSS development projects successful? Mockus
et al. [13] conducted a case study on the Apache Web server and
Mozilla Web browser projects to learn their development process
characteristics; they found that projects based on a relatively small
core of developers (10–15 people) could be geographically
dispersed, yet communicate and function without conflict via a
set of implicit coordination mechanisms (i.e. informal email
exchange). However, when the number of core developers
exceeded this size, other explicit coordination mechanisms (e.g.,
a code ownership policy) had to be adopted. In a similar study,
Huntley [9] used organizational learning to explain the success of
OSS projects; he maintained that it decreased time in fixing bugs.
However, there were significant debugging differences in Apache
and Mozilla, with project maturity as the apparent reason, as
opposed to other factors such as project size and number of
programmers. Debugging data were modeled to fit a learning
curve. Mozilla, an emerging project, was characterized as having
improvements due to learning effects present in their team. Both
these authors pointed out significant differences between the
projects.
Our intent was to extend and refine their work by including a
much larger number of OSS development projects of varying size
(in terms of the number of developers involved) and type (from
simple file management software to complex enterprise software
suite). Specifically, we included 118 OSS projects in our sample. By
focusing on multiple projects of varying size and type, we were
better able to characterize OSS projects. Our study was initiated to
answer the following research questions:
(1)
Are learning effects universally present in OSS projects?
(2)
What are the factors that affect the learning process?
We used the number and percentage of resolved bugs and bug
resolution time to measure learning effects. However, we also
looked at how different project types, number of developers
(project team size) and their experience, and the intensity of
Information & Management 46 (2009) 9–15
ARTICLE INFO
Article history:
Received 3 May 2007
Received in revised form 23 May 2008
Accepted 26 September 2008
Available online 28 November 2008
Keywords:
Virtual organizational learning
Organizational learning curve
Virtual organization
Open source software
Software development
Project performance
ABSTRACT
We studied virt ual organizational learning in open source software (OSS) development projects.
Specifically, our research focused on learning effects of OSS projects and the factors that affect the
learning process. The number and percentage of re solved bugs and bug resolution time of 118
SourceForge.net OSS projects were used to measure the learning effects. Projects were characterized by
project type, number and experience of developers, number of bugs, and bug resolution time. Our results
provided evidence of virtual organizational learning in OSS development projects and support for several
factors as determinants of performance. Team size was a significant predictor, with mid-sized project
teams functioning best. Teams of three to seven developers exhibited the highest efficiency over time and
teams of eight to 15 produced the lowest mean time for bug resolution. Increasing the percentage of bugs
assigned to specific developers or boosting developer participation in other OSS projects also improved
performance. Furtherm ore, project type introduced variability in project team performance.
ß 2008 Elsevier B.V. All rights reserved.
* Corresponding author. Tel.: +1 210 4586337; fax: +1 210 4586305.
E-mail address: yoris.au@utsa.edu (Y.A. Au).
Contents lists available at ScienceDirect
Information & Management
journal homepage: www.elsevier.com/locate/im
0378-7206/$ – see front matter ß 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.im.2008.09.004