Let's first say that I've seen my fair share of physics literature. I've pored over papers, tutorials, etc. with little to show beyond some simple collisions between spheres. What I really wanted to implement was something ideally "not" much more difficult: a rigid body simulation with simple primitives (spheres, boxes, planes) that collide and interact with each other, giving a realistic enough feel to build a game around.
Unfortunately, even this idea is extremely complex and difficult to implement. You would think the math and code would be fairly easy, and it actually is (depending on your math background. A few semesters of college calculus and physics is very helpful, but not required). But programming a physics engine is 40% math and physics and 60% being clever enough to get the computer to do what you want. Floating point errors, inaccuracies in integration and bugs you can't even imagine from the start all mess with your beautiful equations.
And this is where the author, Ian Millington, saves the day with this great book. All throughout, he lays down the challenges of building a physics engine and gives you realistic solutions that are explained very well. You build from the ground up: from a simple particle engine, to a mass aggregate physics engine, all the way to a full blown physics engine. Each chapter he describes what algorithms you need to implement and variations on those algorithms, providing pros and cons for each. I cannot TELL you how fantastic it feels to have an author tell you in an honest tone: here's where this works, here's where it doesn't. Never at any time do you feel that things are being hid behind a curtain of intellect and egomania. And if the author does hide any details, he tells you and with good reason.
In the end, if you plow through all 20 chapters with diligence and fortitude, you're going to end up with a working physics engine. It will be a physics engine capable of a lot of very cool things, but it isn't a finished product. And the author makes it very clear throughout the book that you will not end up with a spit polished physics engine to compete with the likes of Havok. It will be rough on the edges with plenty of room for improvement, but it gets the job done. The author helps you create a working model and leaves you with many options on how to improvement. However, those options require much more advanced coding.
This book provided me with exactly what I was looking for: the know-how and experience to create a physics engine for the everyday game designer. I highly recommend it. If you want to create something commercial, this is a great start, but you're going to have to buy some heavier books on top of this one if you want a very robust, optimized engine.