The scruffy side of requirements engineering

Abstract

In conclusion, the world of the RE is scruffy indeed. Many of the obstacles to success in Requirements Engineering are not amenable to formal methods. Even if some aspects of the problems could be formalized, it may not be worth doing, because the amount of knowledge needed may be too great.

However, research on formal methods for specifying and reasoning about systems and their requirements have had and will continue to have great value for their role in increasing our understanding of system development. (I will not mention any of this work here, since it is the job of those advocating the neat position to toot their own horn.) I would like to see this research focus on the high-leverage problems.

An example of relevant formal work is investigation of deontic and temporal logics. The former is (or at least could be) applied to the RE's task of considering which system components are responsible for which system functions. Temporal logics have mostly been used to specify system behavior but not to describe user requirements in natural terms.

If we want to improve the state of the art in Requirements Engineering, we certainly can't wait for the neats to formalize all the important things. However, we can try to get some leverage on a few important problems. Perhaps the neats and scruffies (who may not be so clearly separable anyway) need to work together to identify the high-payoff problems and applicable formalisms.