PA5: Dependability issues in open-source software
December 2000 - December 2001
Partners: City, Newcastle, Lancaster

Open-source software such as Linux and the Apache web server is becoming increasingly widely used and user reports suggest that this type of software has a high level of dependability. The aim of this activity was to carry out a preliminary investigation of open-source software development with a view to understanding this process, what we could say about system dependability and whether a deeper study would provide to be useful.

This project was carried out during the first year of the DIRC project. Local work at Lancaster was primarily carried out by Mark Rouncefield and Adrian Mackenzie who investigated the sociology of open source development. They studied the development of Cocoon, an open source software system developed in Java that supports XML-based web publishing.

This project presented new challenges for us as our approach to social analysis has always been based on field studies of real organisations. In this case, we developed an approach to studying the virtual organisation involved in the development of Cocoon. This involved studying how the Cocoon web site was used by the developers and analysing the e-mail record that developed during the development process.

Our overall conclusions for this part of the project were that open source development as observed in Cocoon was not really as different from conventional software development as it is sometimes made out to be. Decisions were taken for pragmatic often commercial reasons and the core development team was relatively small. The fact that the code is scrutinised by a range of different people should reduce the number of bugs and hence enhance its dependability. More generally, the conclusions of the overall project were that open-source development was not a single activity but that many different approaches were categorised under this heading. Future work that was identified in the final report includes the need for further investigation of how the open source approach fosters dependability, whether the 'multiple eyeballs' approach to code scrutiny can replace more formal inspections, and the need to understand the relationships between software architecture, group structure and open source development.

For more information on this project, see DIRC web site

The UK MoD have also carried out some investigations of open source development. Their report can be downloaded from here.