The Shape of Code

The book The Mythical Man-Month by Fred Brooks, first published in 1975, continues to be widely referenced, my 1995 edition cites over 250K copies in print. In the past I have found it to be a pleasant, relatively content free, read.

Having spent some time analyzing computing data from the 1960s, I thought it would be interesting to reread Brooks in the light of what I had recently learned. I cannot remember when I last read the book, and only keep a copy to be able to check when others cite it as a source.

Each of the 15 chapters, in the 1975 edition, takes the form of a short five/six page management briefing on some project related topic; chapters start with a picture of some work or art on one page, and a short quote from a famous source occupies the opposite page. The 20th anniversary edition adds four chapters, two of which ‘refire’ Brooks’ 1986 paper introducing the term No Silver Bullet (that no single technlogy will produce an order of magnitude improvement in productivity).

Rereading, I found the 1975 contents to be sufficiently non-specific that my newly acquired knowledge did not change anything. It was a pleasant read, various ideas and some data points are presented, the work of others is covered and cited, a few points are briefly summarised and the chapter ends. The added chapters have a different character than the earlier ones, being more detailed in their discussion and more specific in suggesting outcomes. The ‘No Silver Bullet’ material dismisses some of the various claimed discoveries of a silver bullet.

Why did the book sell so well?

The material is an easy read, and given that no solutions are heavily pushed, there is little to disagree with.

Being involved on a project for the first time can be a confusing experience, and even more experienced people get lost. Brooks can provide solace through his calm presentation of project behaviors as stuff that happens.

What project experience did Brooks have?

Brooks’ PhD thesis The Analytic Design of Automatic Data Processing Systems was completed in 1956, and, aged 25, he joined IBM that year. He was project manager for System/360 from its inception in late 1961 to its launch in April 1964. He managed the development of the operating system OS/360 from February 1964, for a year, before leaving to found the computer science department at the University of North Carolina at Chapel Hill, where he remained.

So Brooks gained a few years of hands-on experience at the start of his career and spent the rest of his life talking about it. A not uncommon career path.

Managing the development of an O/S intended to control a machine containing 16K of memory (i.e., IBM’s System/360 model 30) might not seem like a big deal. Teams of half-a-dozen good people had implemented projects like this since the 1950s. However, large companies create large teams, operating over multiple sites, with every changing requirements and interfaces, changing hardware, all with added input from marketing (IBM was/is a sales-driven organization). I suspect that the actual coding effort was a rounding error, compared to the time spent in meetings and on telephone calls.

Brooks looked after the management, and Gene Amdahl looked after the technical stuff (for lots of details see IBM’s 360 and early 370 systems by Pugh, Johnson, and Palmer).

Brooks was obviously a very capable manager. Did the O/360 project burn him out?