TL;DR
This paper introduces "Rugby," an agile process model designed for continuous delivery in project-based software development. It enhances collaboration and feedback loops by enabling event-based releases, allowing part-time developers to contribute effectively across multiple projects.
What Problem Does This Paper Address?
Traditional agile methodologies like Scrum often assume full-time, dedicated team members and rigid sprint structures, which can be challenging for environments with part-time contributors or projects requiring frequent iteration and feedback. This paper addresses the need for a flexible agile process that accommodates part-time developers, facilitates continuous delivery, and improves communication between development teams and stakeholders, especially in project-based organizations. The authors highlight that in many real-world scenarios, developers are involved in multiple projects, and students in academic settings have limited availability, making strict adherence to full-time agile models difficult.
Methods
The core of Rugby is its agile process model, which integrates continuous delivery and feedback workflows into a software lifecycle. It builds upon concepts from the Unified Process and Scrum, but introduces key adaptations:
- Event-Based Releases: Unlike Scrum's time-based sprint increments, Rugby allows for releases to happen whenever a potentially shippable product increment is ready, triggered by events like customer requests or the availability of feedback. This is facilitated by a continuous delivery pipeline.
- Continuous Feedback Mechanism: Rugby emphasizes frequent interaction between developers and customers. Executable prototypes are used as communication models, allowing for early validation of requirements and features.
- Project-Based Organization: The model is designed for organizations managing multiple projects simultaneously. It utilizes cross-project teams to share knowledge and synchronize efforts.
- Lightweight Roles: Roles like "Team Leader" and "Project Leader" are defined, with the Project Leader acting as an experienced mentor to the Team Leader, fostering a master-apprentice relationship.
- Branching Model: A simplified version of the Gitflow branching model is adopted to manage feature development and integration, allowing for the creation of multiple prototypes.
Datasets
Not discussed in the paper. The evaluation relies on data collected from university capstone courses.
Key Insights and Ideas
- Executable Prototypes as Communication Tools: The paper strongly advocates for using executable prototypes not just as code, but as a primary means of communication. This significantly reduces the time and effort required for status reports and discussions, especially in multi-project environments.
- Flexibility in Release Cadence: Moving beyond fixed sprint-end releases, Rugby's event-based approach allows for more agile responses to feedback and changing requirements, leading to faster iteration cycles.
- Bridging Gaps: Continuous delivery in Rugby is shown to bridge not only the gap between development and operations but also between developers and customers/users, fostering better collaboration and understanding.
- Empowering Part-Time Developers: The model's structure, with its emphasis on clear communication channels and event-based workflows, makes it more amenable to part-time contributors and students with limited availability.
Key Results
The evaluation, conducted in two large university software engineering capstone courses with up to 100 participants and 10 simultaneous projects, yielded promising results:
- Increased Frequency and Quality of Interaction: Rugby demonstrably increased the frequency and quality of interaction between developers and customers.
- Improved Coordination: In 2013, using executable prototypes in management meetings led to significantly less time spent on status reporting and better understanding of project progress compared to 2012.
- Higher Build Success Rate: In 2013, 75% of commits led to a successful build, compared to only 15% in 2012, indicating improved integration practices.
- Increased Prototype Delivery: The number of executable prototypes delivered to customers was three times higher in 2013, attributed to the build server being available from the start of the project.
- User Feedback: In 2013, crash reports and structured feedback mechanisms generated 139 reports, indicating active user engagement.
Limitations and Future Work
- Generalizability: The authors acknowledge that the findings are based on a university setting and may not directly translate to all industrial environments. While they believe some findings are applicable to industry, further validation is needed.
- Selection Bias: The online questionnaire had a response rate of 41 out of 90 participants, and the authors note that some teams used the workflows more than others due to experience or customer requests. They mitigated this by analyzing results on a team basis and observing no significant deviations.
- Future Work: The authors plan to investigate Rugby in an industrial project setting in 2014 to further validate its applicability in more realistic circumstances.
Summary updated