Coordination Dynamics in Free/Libre Open Source Software Development

Andrea Wiggins

Syracuse University iSchool

3 June, 2009

Introduction

Coordination comprises activities through which interdependencies are managed

Key challenge for decentralized, independent virtual work, but FLOSS seems to do this well A variety of mechanisms are used to

reduce interdependencies

Motivation

Coordination in virtual teams poses different challenges from face-to-face

Coordination mechanisms are not applied in isolation, but employed in combination as coordination strategies May represent common good solutions to

general organizing problems in FLOSS Stages of development likely to affect

dynamics of coordination strategies

General Research Question

How does project growth affect the social and technical structuring of work through coordination strategies? Project growth has multiple aspects: code

& community Coordination is achieved through multiple

social and technical mechanisms

Specific Research Question

How do the dynamics of the relationship between the size of the core committer group and the size of the code base affect technical coordination through code modularity and social coordination through self-assignment to tasks in community-based FLOSS projects? Growth: core committer group and code base size Coordination: code modularity, self-assignment Process: dynamics Population: community-based FLOSS projects

Conceptual Framework

P1: As size of code base and core committer group increase, code modularity increases.

P2: As size of code base and core committer group increase, self-assignment to tasks increases.

P3: As code modularity increases, self-assignment to tasks increases.

Methodology

Longitudinal multiple case study methodology using mixed methods Correlational analysis on archival data Qualitative narratives of coordination

dynamics, drawn from content analysis Replication and extension of Crowston et al.

2005

Effort required for manual content analysis constrains sample size

Case Selection

Community-based projects Same type of software

Moderately complex One primary package

Data available in repositories

No continuous release cycle projectsMinimum 10 developers, 18 months of

data, 3 releases

Data

Observation sampling of developer email lists 3 weeks before and 1 week after each

release: minimum 12 weeks of email/case Email is a primary communication venue,

can apply an established coding schema

Project statistics - FLOSSmole & SRDACode metrics - FLOSSmetrics

Analysis

Content analysis according to established schema, compare findings to prior results

Test correlational measures of scale of coordination effort

Qualitative narratives of coordination dynamics: contextualize simple measures in evolving work practices

Validity

Case selection bias improved by purposive sampling for success

Potential for measurement error from project statistics and code metrics Simplistic but direct operationalizations

Limited sample restricts generalizability Plan to test semi-automated coding to

increase scale of content analysis

Expected Contributions

Advance process theory to explain coordination strategies as an outcome of scale and interdependency of work

Reproduce and extend prior work: Implement dynamic analysis of multiple coordination mechanisms

Evaluate claims of relationship between community size, code structure, work

Outstanding Issues

Scale of content analysis Semi-automated coding may help

Case selection - comparable softwareOther coordination mechanisms

Cannot control for them in case selection May emerge as important factors

Code modularity/complexity Limitation of single package software

Completed activities

Candidacy exams: December 4, 2009Proposal defense: no earlier than May 2010Different dissertation topic is likely

Massive virtual collaboration in citizen science This study will be post-PhD work

Thanks!

awiggins@syr.edu

www.andreawiggins.com

floss.syr.edu

www.flosshub.org