The Mythical Man-Month, in detail
Frederick Brooks managed the development of OS/360, IBM's operating system for the System/360 mainframe, in the early 1960s. It was one of the largest software projects undertaken to that point, and it was very late, very over budget, and enormously complex. The Mythical Man-Month is his account of what he learned from that experience — an account specific enough to still be accurate and influential fifty years later.
The title essay establishes the book's central argument: adding manpower to a late software project makes it later. This is not intuitively obvious. If two programmers can write a program in six months, ten programmers should be able to write it in one month — but they cannot. The reason is that software development is a complex coordination task, not a simple parallel one. Adding people requires training, communication overhead, and the partitioning of the work in ways that may not be possible given the dependencies in the design. Brooks's formulation of this has become Brooks's Law: adding manpower to a late software project makes it later.
The book is full of insights that remain accurate. The second-system effect predicts that engineers allowed to build their second system after succeeding with their first will over-engineer it disastrously, freed from the constraints that made the first system tractable. The surgical team model argues that large programming teams should be organized around a small number of very skilled people supported by specialists, rather than assembled from a large pool of interchangeable workers. The essay on conceptual integrity argues that every large system needs a coherent underlying vision, which requires concentrated authorship rather than design by committee.
The anniversary edition includes "No Silver Bullet," Brooks's famous 1986 essay arguing that there is no single development methodology, tool, or technique that will produce an order-of-magnitude improvement in software productivity, because the essential difficulties of software — the complexity of what it must represent — are irreducible. This essay remains one of the most important pieces of thinking about why software is hard, and why the promise of each new methodology to solve the problem has been consistently over-stated. The book as a whole is essential reading for anyone involved in managing or building software systems.
The big ideas
- 1.
Brooks's Law: adding manpower to a late software project makes it later. Communication overhead and training costs swamp the productivity gain from additional workers.
- 2.
The second-system effect: engineers building their second major system tend to over-build it, adding all the features they held back from the first. Restraint on the second project requires conscious effort.
- 3.
Conceptual integrity — a coherent underlying design vision — is the most important quality of a large system and requires concentrated authorship, not design by committee.
