A Study of Maintainability in Evolving Open-Source Software

TL;DR

This study evaluates the maintainability of long-term evolving open-source Java GUI applications using multiple models, analyzing trends, drivers of change, and distribution of issues over more than a decade.

Contribution

It provides a comprehensive, multi-model analysis of maintainability evolution in open-source software, with cross-application comparisons and detailed source code insights.

Findings

Maintainability varies significantly over time and between applications.

Software size influences maintainability trends.

Common patterns and application-specific issues were identified.

Abstract

Our study is focused on an evaluation of the maintainability characteristic in the context of the long-term evolution of open-source software. According to well established software quality models such as the ISO 9126 and the more recent ISO 25010, maintainability remains among key quality characteristics alongside performance, security and reliability. To achieve our objective, we selected three complex, widely used target applications for which access to their entire development history and source code was available. To enable cross-application comparison, we restricted our selection to GUI-driven software developed on the Java platform. We focused our examination on released versions, resulting in 111 software releases included in our case study. These covered more than 10 years of development for each of the applications. For each version, we determined its maintainability using three distinct quantitative models of varying complexity. We examined the relation between software size and maintainability and studied the main drivers of important changes to software maintainability. We contextualized our findings using manual source code examination. We also carried out a finer grained evaluation at package level to determine the distribution of maintainability issues within application source code. Finally, we provided a cross-application analysis in order to identify common as well as application-specific patterns.