Computer system architects and designers may find that the value of this thought-provoking and difficult paper lies outside the explicit comparisons made by the author and more in the overall direction in which he points.
What is new here is the juxtaposition of computer system design with the design theory of a practicing architect, Christopher Alexander [1]. The inspection of the activity undertaken by the architect of physical buildings has implications for the architect of computer systems. Also valuable is Olerup’s emphasis, following Alexander, on the way the system (whether it is a software product or a building) relates to its context.
The strongest part of Olerup’s paper is his account of Alexander’s approach to architectural problem solving as designing the satisfactory fit of form and context (p. 220). Design problems show up as a “misfit” (p. 221) between some structure (whether it is a form of life, a dwelling or household utensil, or a logical entity such as an automated construct) and the context in which it is to function. Presumably, the entire apparatus of diagrams, directed graphs, and their automated instrumentation will be deployed at this point to perform the business requirements analysis and specification. The form can then be synthesized from the detailed features that will satisfy the goals of the system users and will fit the context in which the form is implemented.
The paper has three weaknesses, two of which can be easily remedied. The third, however, is likely to resist a quick fix.
First, the model of information systems development on which Olerup relies dates back to the late l970s. It could be easily updated by incorporating additional material on automated CASE and design methods and tools.
Second, the author does not really define the distinction between “technical” and “procedural” rationality. Granted that the models provided by mechanical systems and machines have limitations, what is the alternative? Organic systems? Meta-models? Logical constructs? Some hybrid?
Finally, the concepts of “goodness of fit” and “misfit,” which are drawn by Alexander from gestalt psychology and are inherited by Olerup, are notoriously difficult to define. These thinkers, including Olerup, have to fall back on a negative definition--cognitive dissonance is experienced when form and context do not harmonize. This definition implies that designers will continue to rely on art, soft methodology, or intuition at certain crucial phases of the design effort. Still, this vagueness is consistent with the fact that much more can be done to automate system architecture design. In this regard, Olerup’s thoughtful comparisons are likely to provoke further useful dialogue.