general

I learned both network programming and threaded programming (pthreads) from this book (the 1998 2nd edition), and it was a great teacher. I used it to design and implement a client server protocol for a research project involving 10 machines and about 15 communicating processes. One of the things I really appreciate about the book is the detailed examples.

After the concepts (a server 'listens' for a connection; a client initiates a connection) and basics of packets and sockets, the book presents a simple tcp server and client -- the server simply echos back whatever the client sends. Next, it discusses what can go wrong with the simple version, and presents an improved echo server that behave better when the client crashes (and likewise an improved client).

The book continues to improve on the basic client/server, including address resolution, and servers that handle multiple clients using forks, threads, non-blocking I/O. This is all I needed to learn from the book, and it's all in the first half of the book.

Particularly useful is Chapter 27, Client-Server Design Alternatives. To me it alone was worth the price of the book. Here the book discusses concurrent servers with three basic architectures:

(1) non-blocking I/O, no threads or forks (advantage: full control of resource allocation; disadvantag: complexity);

(2) spawn a thread or fork for each client (simplest implementation; potential problem of too many children);

(3) servers that pre-allocate a pool of threads or forks (a happy medium; faster).

Other chapters discuss broadcasting, multicasting, out of band data, routing sockets, and raw sockets; all topics I hope to learn some day. Again, this is a great book. The 3rd edition brings it up to date for IPV6 and numerous small improvements.

This will become the new classic of Unix Network Programming. When I first heard that there was a new edition, I was afraid that the classic would be destroyed. Although it is quite different and much of it has been done by "et al", it turns out to be quite good.

Initially I did a quick scan and immediately saw some things missing (like FTP). However, after reading the book and realizing how infrequently I had referenced the missing parts in the first book, I was much happier. With the additional IPv6 information and newer technologies, this more than made up for the deletions (from volume one).

The text is easy to read and flows logically. What really impressed me was the depth of the expertise. For me, there had been some stumbling blocks in TCP and UDP development early on (especially nuances of UDP) that took a fair amount of time to figure out. I was pleasantly surprised to see many of those difficulties laid out cleanly for anyone to figure out in a matter of minutes. That alone is worth the price of the book.

The index is extensive and helpful. The questions at the end of each chapter are limited, but enough to be useful in a college course.

The printing is high quality. The paper is high quality (thin, but strong so that the 1000 page book is not huge or heavy). The printing takes up much of the page (i.e., not two-inch borders like some books). The binding lies flat without separating the pages. Definitely a quality printing.

Yes, a new classic.

I bought this book about a month ago and used it very extensively for my UNIX programming project. While this book is great and very comprehensive, I did find code examples very confusing. As coping a variable from the socket struct into variable and then from this variable into custom structure, then manipulating this custom structure and finally printing. It would be better to use as less transformation as possible. Because it makes very hard to find what this variable has and where it came from. At the end all I'm looking for is simple but good example of how to use sockets...etc, it's not a C programming book.