Teachers already attempt to structure their courses around certain organizational principles. Two of the most popular approaches recommend a ``top-down'' design methodology and advocate the study of computer science through the study of algorithms.
The principle of top-down design says that a problem should be decomposed into several fragments, each of which is easier to solve; each sub-problem is then solved using the same principle; finally, the solutions to the fragments are combined to produce a solution to the complete problem.
Top-down design is a reasonable idea but it suffers from several shortcomings. First, it is too abstract: It provides no concrete guidance on how to perform the decomposition, or how to combine the solutions to the fragments. Second, it runs counter to actual program development, which is partially top-down and partially bottom-up. Finally, by emphasizing the construction of systems in a monolithic fashion, it discourages experimentation, and offers no hints when the final program has errors.
The algorithmic approach suggests that, since computer science centers around problem-solving, the best way to teach the subject is to present a variety of problems and teach algorithms that solve them.
In essence, this approach could be orthogonal to ours. Unfortunately, the emphasis on devising clever solutions neglects to teach students how to structure the ensuing solutions. As a result, even after students have designed an algorithm, they have difficulty implementing it in a correct and natural fashion. We believe instead that students should concentrate on the elements of design, and use systematic techniques to derive algorithms. This not only gives them a solid foundation, it also improves their understanding of the algorithm, making it more likely they will recognize superficially different domains where the same solution can be used.
PLT / scheme@cs.rice.edu
Last modified at 10:49:17 CST on Monday, November 10, 1997