Agile management of uncertain requirements via generalizations : a case study

A major justification for the agile class of software processes is that customer requirements often change substantially during development, with the result that developers must work with a ever-changing specification. The best way of dealing with this, it is claimed, is a process including adaptive and iterative steps allowing the code under development to change to meet an evolving requirement. In this paper, we use quantitative process data to argue that the deliberate development of an evolvable design, based upon extensions and variations to the requirements, and the generality derivable from the implemented system, may provide agility within a more traditional development, showing that the two concepts are interchangeable, to some extent. In doing so, the paper draws upon our joint work on evolvable systems, and upon a concept developed by one of us in the 1980's, "The Contraction of Generality", to illustrate the alternatives.