The Mythical Man-Month: Essays on Software Engineering, Anniversary Edition [Paperback]

You will realize that long before maybe you were even born, other people working at places like IBM had already experienced those problems and quandries. And found working solutions to them which are as valid today as they were 30 years ago.

The suggestions in this book will help you think better and better manage yourself, and be more productive and less wasteful with your time and energy. In short, you will do more with less.

Some of Brooks insights and generalizations are:

The Mythical Man-Month:
Assigning more programmers to a project running behind schedule, may make it even more late.

The Second-System Effect:
The second system an engineer designs is the most bloated system she will EVER design.

Conceptual Integrity:
To retain conceptual integrity and thereby user-friendliness, a system must have a single architect (or a small system architecture team), completely separate from the implementation team.

The Manual:
The chief architect should produce detailed written specifications for the system in the form of the manual, which leaves no ambiguities about any part of the system and completely specifies the external spcifications of the system i.e. what the user sees.

Pilot Plant:
When designing a new kind of system, a team should factor in the fact that they will have to throw away the first system that is built since this first system will teach them how to build the system. The system will then be completely redesigned using the newly acquired insights during building of the first system. This second system will be smarter and should be the one delivered to the customer.

Formal Documents:
Every project manager must create a roadmap in the form of formal documents which specifies milestones precisely and things like who is going to do what and when and at what cost.

Communication:
In order to avoid disaster, all the teams working on a project, such as the architecture and implementation teams, should stay in contact with each other in as many ways as possible and not guess or assume anything about the other. Ask whenever there's a doubt. NEVER assume anything.

Code Freeze and System Versioning:
No customer ever fully knows what she wants from the system she wants you to build. As the system begins to come to life, and the customer interacts with it, he understands more and more what he really wants from the system and consequently asks for changes. These changes should of course be accomodated but only upto a certain date, after which the code is frozen. All requests for more changes will have to wait until the NEXT version of the system. If you keep making changes to the system endlessly, it may NEVER get finished.

Specialized Tools:
Every team should have a designated tool maker who makes tools for the entire team, instead of all individuals developing and using their private tools that no one else understands.

No silver bullet:
There is no single strategy, technique or trick that will exponentially raise the productivity of programmers.

Lets say you are late and another programmer is assigned to help you out - you can simply point to this book and explain how adding more programmers to a late project will just make it later. If that one doesn't fit. you can surely find another one that does.

If you liked "Catcher in the Rye" or "The Peter Principle" you may like this book too.

Software management is hard - mostly because there is a great deal of variation in the talent and productivity of computer programmers.

This book is a fun read - and food for thought. And in its defense I must admit it has changed the way I think about large software projects.

But sadly, beyond the fun quotes and maxims (which often contradict each other) there is not much to help you get the job done.

But software engineering has advanced a lot since then, even if the software industry hasn't. For example, Brooks' sole team-level improvement is the suggestion to use Harlan Mills' chief programmer teams, while many such improvements have been found since then. And Brooks entirely ignores the main defect of the chief programmer team---the difficulty of finding chief programmers!

(As an aside, a chief programmer team works fine now with a chief programmer, a college grad, and modern tools. Code ought to be written so a college grad can maintain it, and this approach helps ensure that. The college grad can also flesh out test cases and support in other ways. But there's still the problem of finding the chief programmer...)

Brooks approach is generally, "We did that wrong. We should have done it this way, for these logical reasons." But there are often several solutions to a problem, all having logical reasons. Empirical data is needed to choose between them. Brooks rarely mentions alternate solutions, and almost never offers emperical data.

A far more valuable book is Steve McConnell's "Rapid Development". This well-researched and organized book quotes data to confirm problems, discusses solutions with associated emperical data, and recommends solutions.