A Disturbing Result on the Knowledge Used during Software Maintenance

As part of a long term project aiming at empoweringsoftware maintainers with knowledge based tools, we conductedan empirical study on the knowledge they use whendoing maintenance. The study was intended as a preliminarywork to get initial data on the type on knowledgeused and the importance of each type of knowledge. Forexample, it is commonly assumed that application domainknowledge is important when doing maintenance. Howevernobody can tell exactly how much more important it is thanComputer Science knowledge.We monitored six software engineers in two different organizationsand analyzed the knowledge they used in theirmaintenance activity. In this paper we present and discusssome results and propose research directions from these.Our results include: the fact that software engineers rarely"search for" some new knowledge and rather work fromwhat they already know; or the acknowledgment of a variationof the knowledge used depending on the organizationone is working in. The most disturbing result is that, contraryto what is commonly assumed, the software engineerswe studied seemed to make little use of application domainknowledge.