Configuration is a design task that is the target of much AI research. It is a comparatively tractable design task and thus
can be completely automated in a knowledge based system (KBS). Indeed the earliest commercially successful KBS was XCON, a
rule-based system for configuring VAX computers developed by Digital Equipment Corporation.
KBS technology has moved on and the dominant techniques used in configuration now are constraint handling techniques. However
expertise in configuration is evidently experience based and there has also been considerable research on using CBR for configuration.
Normally, configuration problems are comparatively closed problems and so can be completely modeled in a KBS. For this reason,
a CBR system for configuration can be a completely automated system rather than an interactive assistant, as is the case in
more difficult open design tasks. It will be seen later in this chapter that this is a defining characteristic of CBR systems
for configuration.
We believe that this characteristic of configuration, as a design problem that can be completely automated in CBR, tells us
something about CBR. If we view adaptation as a configuration task then it too can be automated. We pursue this idea in the
second part of this chapter where we analyze two case adaptation problems as configuration tasks. We show that this adaptation- as-configuration perspective provides insights into what can and cannot be achieved using automatic adaptation in CBR.