Assessing the Relevance of Identifier Names in a Legacy Software System

Reverse engineering is a difficult task even for humans. When trying to provide tools to assist in this task, one should try to take advantage of all possible sources of information. Informal sources, like naming conventions, are more abstract than the code, thus helping to bridge the gap between code and design. On the other hand, there is no certainty that they actually represent the current state of the system. Some researchers have used these informal sources of information, relying on an assumption that if types (or variables, or functions) have the same name then they represent the same thing. However none of these researchers have actually tried to establish to what extent this assumption is valid. This paper provides a framework to study this issue. We will define what it means to have a "reliable naming convention", how this can be tested and under what conditions. We present examples from the particular legacy software system we are studying as well as from the literature. ..