Every rating system is only as good as its explanation. If you don't know how I work out my ratings, they're not really good enough for you to make up your own mind. You shouldn't buy a book just because I rated it good. But it might be a good idea to buy and read a book if I rated it good and you agree with my argument why I rated it good. Some parts of that argument are clear from the text. Other parts are applicable to all books, and they're described here.
The two most important parts of my rating are:
More detailed parts of my ratings are:
The standard book about all meta-topics around software development in general, not only applicable to UNIX. Covers project organization (directory structure, essential files), community communication, important requirements, domain specific languages and more. A MUST have for all programmers, not only UNIX programmers.
If there is one single book that a true geek's bookshelf must not be missing, then it's this one. I've only recently heard of this book, and actually this book was the book that made me starting to write book reviews.
The books sole content is a table of one million random numbers of high quality plus a table of a hundret thousand normal deviates.
An interesting book about artificial intelligence for game developers describing algorithms for weak AI. Interesting, but don't expect too much from it. (Not yet finnished)