The focus on using expert knowledge to provide a hierarchal system for design makes this a nice example of this technique. Rather than go on about their own ideas about what design is, they limit their problem to a search space in which designs have same functionality and known parameters. It complements the DSPL paper very well.

      PRIDE can be compared to an optimization system, like Dominic. While it does not explicitly handle optimization, it can be expressed as goals and design methods, for example. This might make the system more flexible or just cause extra work for the author of a particular problem solving plan. It would be interesting to compare results with a system, such as Dominic, that uses hill climbing instead of hierarchical search.

      By allowing the ordering of parameters to be specified as knowledge PRIDE provides an interesting mechanism to balance between efficiency and initial success rate. I think this might be only system we have looked at so far that represents this kind of ordering as knowledge.

      Although goals seem to be just another construct in the hierarchical design process, the idea of goal based design decisions is very powerful. Many modern uses of goal based reasoning can be found in MIT's Media Lab. It is amazing how hot topics of discussion from the 1980's are still on the cutting edge of Computer Science.

      The description of advice for modification upon failure was a little confusing at first. After the problem solving section it was more clear but the use of terms and presentation made this feature seem mysterious. Overall, the advice mechanism seemed very similar to DSPL's suggestion/message passing.

      The authors speak of how to address tightly coupled variables though not in an example problem solver. I would have liked an example. Also, the multiple contexts mechanism might help address this but was not mentioned as a possible solution to manipulating the values of tightly couple variables.

      It would have been nice to see an example of a design problem from start to finish. The closest the paper came to outlining the operation of the whole system seems to be Figure 9-2. Goal Hierarchy. The statement about the DSPL system not searching the design space as thoroughly as PRIDE seems incorrect since DSPL could search the space completely.