This book is unlikely to be of interest, unless you’re interested in data processing programs written in COBOL. However, I was interested in optimisation and the origin of:
1. Don’t do it.
2. Don’t do it yet.
So I was satisfied with that at least. There are ideas here which have survived rather better than the examples, such as keeping programs simple, avoiding GO TO where possible and making each component responsible for validating the data it passes on. But much of the book is firmly rooted in the sort of data processing systems it was written for, and I struggled to relate many of the design techniques, particularly program inversion, to the way I think about software today. I did enjoy the occasional bit of personality, especially “A Moral Tale” and warnings to “resist this satanic temptation”.
The exception was the chapter on optimisation, which was by far the most interesting part of the book. Jackson’s two rules are wonderfully simple: “Don’t do it” and “Don’t do it yet.” His argument is that optimisation needs a positive, quantified justification and should only be attempted once you have a clear, simple and unoptimised solution. Making software faster or smaller can also make it harder to understand, maintain and change, and more prone to error. Nearly fifty years later, that still feels entirely relevant. I also liked his point that “optimization” is really the wrong word: making something faster or smaller does not necessarily make it better.
There was one final curiosity about my copy. It was printed in Great Britain by Page Bros (Norwich) Ltd in 1981, in my home town. Despite that, more than forty years later I had to have my copy shipped back across the Atlantic from the US. That was probably more interesting to me than quite a lot of the book. Principles of Program Design is an interesting historical artifact, and I’m glad I found the source of the optimisation rules I was looking for, but beyond that there wasn’t much here for me.
Principles of Program Design
M. A. Jackson
ISBN: 0 12 379050 6
Comments
Post a Comment