Alistair Sutcliffe
Springer, 2002
Alistair Sutcliffe
Springer, 2002
ISBN: 1852335173 paper.
Buy it from Amazon.com
Buy it from Amazon.co.uk
This compact but tightly-packed little book is "primarily a research text for graduate courses" but is clearly also of interest to practitioners who want to go a little further than usual into "why RE is difficult and how human understanding can cause the problems we observe in getting requirements right".
With the warning in the preface that this book is in part a review of Sutcliffe's own research in mind, this is an interesting book in several ways. There is a strong emphasis on the human aspect, both from the point of view of the usability of systems (thinking about user goals, tasks, and dialogue) and from the point of view of human cognition (thinking about why people make mistakes, what motivates them, how people reason, and the like).
It is almost bizarre how little cross-over there has been between, say, the HCI and the RE communities. HCI is powerfully influenced by psychological technique and experimental findings, indeed by the scientific method in general. RE remains, despite its boasts about being problem- and stakeholder- rather than system-centred, distinctly short on psychological and philosophical insight, and not especially long on concrete theory.
Sutcliffe gently explores the fuzzy boundary area between business analysis, systems analysis, human interface design, human behaviour (both individual and social) and system specification. Somewhere in the middle there is an important area of research and practice -- ours. For example, how often do people try to define business needs without thinking about power? You only need one powerful but unexamined adversary to doom your project; it is a bigger risk than almost anything else, yet the word 'politics' is virtually taboo (and it is indeed very dangerous).
Part of the problem of bridging gaps between different communities of researchers and practitioners alike is their use of language. In this, Sutcliffe is like the rest of us probably more partisan than he realizes. For instance, he talks about "processes for discovering and refining requirements", which as it happens I find a clear and well-chosen phrase; but other people talk about capturing or eliciting or defining, and the objects of these processes are called requirements, or knowledge, or business rules, or use cases, or whatever -- with the result that minds often never meet, and the same real-world problems (e.g. 'people don't know what they want!') are reported anew time and again. In fact, the index does not contain 'capture' or 'elicitation' (or HCI for that matter).
What this book does very well is to show the state of the art by documenting the history and progress of RE over the last 20 years or so. The knowledge demonstrated is impressive and the references cover ten pages of small type, going back at least to 1976 and a paper by Bell and Thayer charmingly entitled "Software Requirements: Are they really a problem?" which was a title you could get away with back then.
Sutcliffe is also very good on models: not giving screeds of boring definitions of what their syntax is, but showing why you want to use models of different and perhaps to many practitioners unusual kinds -- semantic networks, scripts, category hierarchies, goal trees, i* dependency models, trade-off matrices, dialogue patterns for event-driven analysis, decision trees, naive rich-picture-like 'domain model diagrams' (the example is a ship-board emergency management system), design rationale graphs, even dataflow diagrams (for mapping the policy-goal route for requirements discovery). Each one is shown to be a useful practical tool for a specific RE task. This strength is one of the things that help to make the book valuable for advanced practitioners.
If you want a book that will give you well-founded cause for concern about how people are currently researching and practising RE, and that on the way teaches you how to put together the best from philosophy, linguistics, psychology and other serious chunks of human knowledge, you should order a copy of Sutcliffe.
© Ian Alexander 2002