Lessons learned from a software process modeling system
Herb Krasner, Jim Terrel, Adam Linehan, Paul G. Arnold, William H. Ett · Communications of the ACM · 1992
LessonsLearned from a Software Process Modeling System ompetitive pressures on the software industry from home and abroad have encouraged organizations to examine the effectiveness of their software development and evolution processes.This examination is occurring within the context of establishing performance baselines, setting improvement goals, explicitly defining the normally implicit processes and measuring progress toward improvement goals.A typical objective of a company engaged in a software improvement initiative focused on achieving higher capability maturity levels is to document the current software process (i.e., "as is" baseline) and define one or more ideal processes (i.e., "to be" goal) to strive for.Software process modeling is the technique used to define and analyze the significant aspects of these processes.It is done in order to facilitate human understanding of the process, support process management and continuous process improvement and facilitate automated guidance and enaction of resulting process definitions.From observations gathered during rigorous studies of actual software development situations at the Microelectronics and Computer Technology Consortium (MCC), Curtis and Krasner et al. concluded that processes such as opportunistic design, learning, technical communication, negotiation, and customer interaction are among those which are most crucial to project success but are poorly provided for in most existing process models [7].As a result of the detailed field study of large software projects, Curtis, Krasner, and Iscoe concluded that having an agreed-upon and commonly shared software development process model is a major factor in a project's software development effectiveness [6].Unfortunately, a recent analysis of the data collected by the Carnegie Mellon University Software Engineering Institute (SEI) indicates that the vast majority of software development organizations in the U.S. do not possess and/or use a defined,