This book is an excellent, practical introduction to game physics engine developers. The goals is to take a good programmer [and 3D engine/game developer] with little or no physics experience, and help him upgrade his 3D engine from "no physics" to "modest physics". Thus the book covers the most fundamental and important types of physics for most games, but omits most niche or highly advanced [special purpose] physics.
The great strength of this book is the near absense of "arm waving". The types of physics this book covers are covered more-or-less completely, and implemented in the code that can be downloaded and inspected.
In most 3D books, and especially "physics engine" books, the author discusses each issue more-or-less independently, and in varying levels of detail (for his convenience), then leaves the reader to figure out how to effectively implement the details of the technique, and worse, to encounter and deal with the many practical tradeoffs and gotchas on his own.
In contrast, this author discusses each issue in exactly as much depth and detail as necessary to implement code that works, and deals with practical tradeoffs, gotchas, and integration with the rest of the engine. This makes the book far more helpful for programmers who are not scientists or not very familiar with newtonian-style physics.
Frankly, I'd love for this book to have a "second half" that revisits each level at a higher, more advanced and/or more abstract level, plus discusses many niche/specialty types of physics for game engines. However, the author and publisher leave the Eberly "game physics" book to satisfy that desire. The Eberly game physics book is also excellent, but suffers from the common "weaknesses" that I mention this book does not suffer from. On the other hand, "weakness" is a relative term --- the Eberly book may be written in a perfectly fine manner for scientists, physicists, and especially math wizards (or at least highly comfortable with advanced mathematics).
This book does have a few weaknesses.
It fills far too many lines and pages with same source-code. The problem is not that he includes source-code, that much is fine. The problem is, most of the source-code he includes in the text of the book contains lengthy comments that simply repeat what we just read in the body of the text. This wastes a LOT of space, pushes related paragraphs onto separate pages where they are more difficult to glance back and forth to compare, and makes the few lines of actual code difficult to locate visually between the lengthy comment blocks. He should omit most if not all these comments, then the source-code would make a substantial contribution to the book, rather than mostly just be annoying.
I found a few errors in the book, but so far have not found errata anywhere (for example, on the [...] website).
I also have not yet found a forum dedicated to this book and the physics engine it implements. This would make it a lot easier for readers to get quick, easy answers to questions, confusions, errata, etc.
While the author carefully explained his notation, he would have been better served to adopt the simplest notation possible for this book aimed at the minimally math-savvy audience the book is otherwise so perfectly written for. I mean, is there any reason we must read f = mp (with two dots over the p) instead of f = ma ??? It is as-if "v" for "velocity" and "a" for "acceleration" are x-rated words, and this is a PG book... or something. Elsewhere the author wisely adopted the most clear, simple ways to express his ideas and the math. He should have stuck to that principle. Alternatively, almost everywhere he had plenty of room to state each equation in both forms. This would have been a great service to less math-savvy readers, as seeing equivalent forms side-by-side repeatly is a great way to become comfortable with more advanced symbology and formulation.
Bottom line: The strengths of this book are far greater and more precious and rare than the weaknesses. Everyone trying to implement a physics engine should buy this book. However, physics engines are chock full of subtle issues, niche fields, gotchas and quirks, anyone who actually implements one should get every recent (less than ~7 years old) book and PDF, and read everything, and keep them around for reference, comparisions, and varied presentation.