First of all, this book is no UNPv1. If you want to learn network programming for UNIX/Linux, there is nothing that could replace UNPv1 (Stevens "Unix Network Programming" 2nd ed., vol. 1). [CORRECTION: actually Comer's book(s) are comparable to those of Stevens]
Now back to the book reviewed. I bought it because it seemed to elaborate a bit more on a subject covered only briefly by Stevens: concurrent network I/O in real-world conditions. Well, it did have a special section on the subject but it didn't provide any particular information I was looking for:
- analysis of different I/O methods' behavior under different load conditions
- throughput analysis
- applicability of different methods in different UNIX flavours (Oops, this is a book for Linux! Still, it would be nice if the author at least mentioned the portability issues)
- while the examples were nice and clear, it would make sense to create an I/O abstraction layer that would act as a (de)multiplexor and to change only this layer for every I/O paradigm.
So, once we get past the section on real-world network I/O issues, what we are left with is a decent yet not perfect book on network programming. It will teach you how to develop network applications for Linux but it won't provide the same depth UNPv1 does.
Conclusion: Read it if you want to quickly start developing network apps and you don't have time for UNPv1. Otherwise go for UNPv1. Still, it is worthwhile to skim through the sections on throughput and performance.