Title:
A Decision-Making Framework for Software Maintenance and Evolution
Abstract
In this talk I will outline a research framework for making software maintenance and evolution decisions. Currently most decision-makers (project managers and architects) make software maintenance decisions on an ad hoc basis. They do this because they have no foundation of data on which to base refactoring decisions and they do not have the tools to make rational cost/benefit-based decisions. As a result software inevitably decays over time. To address this problem I will introduce a decision-making framework to support rational decision-making, explain how it works and how we populate it with software maintenance data. The core of this approach rests on two abilities: 1) the ability to correlate measures of software complexity with measures of maintenance effort, and 2) the ability to efficiently find potential software maintenance problems in a code base. In this talk I will primarily discuss the first point--correlating measures of software complexity to measures of effort. I will present our new approach of relating file metrics to maintenance effort measures, which attempts to correlate the relation between changes in file metrics over multiple releases and changes in the maintenance effort spent on these files. And I will present our empirical results from extracting and correlating effort and complexity metrics from a number of open source project. Finally I will conclude by discussing the overall implications of this framework for software maintenance and evolution.