The Linux Network with CDROM [Paperback]

Moral: Beginner with Linux and you want to get into networking, then buy this book. Experienced with Linux and you can learn easily from the avaliable HOWTOs, then shy away from this book.

In fact, this line really should have read, "We have not tested the descriptions in this book to ensure their accuracy; this book will contain many errors."

Notice: the book is full of mistakes and is dangerously misleading in more than a few areas. I would have given it 5 stars had I reviewed it early on, but as I delved deeper into the reading, relying on it heavily as a reference, it became evident that many of the explanations were misleading, ill-referenced, erroneous, and/or downright incorrect.

Even great authors are not exempt from the responsibility of proofreading their work. It seems this book missed that phase altogether. In some sections, you cannot read a single page without finding an error. (Ex: DNS). Many of the errors, unfortunately, are not trivial:

- The definition of Input in the IP masquerading section is wrong. "Input" in and of itself does not specify direction. Input can be from the Internet as well as from the Intranet. Ironically, the back-to-back definition for Output IS correct--small example of many such inconsistencies throughout the book.

- The explanation on setting up two ethernet cards, aside from being inconsistent, fails to mention something very important. Passing arguments to the kernel through lilo.conf only works if the NIC's driver is compiled directly into the kernel. If the NIC is being loaded as a module, then parameters need to be passed through conf.modules NOT lilo.conf. In chapter 3 the user is instructed to compile the Ethernet driver directly into the kernel, granted, but not a single thing is said about the option and ramifications of compiling Ethernet drivers as modules instead, a very common and efficient practice for many Linux users.

- Where descriptions are brief, the author points the reader to the end of the chapter for "additional reference." When the reader flips to the end of the chapter, what do you know? There are no references there.

- The part on Configuring Masquerading (p.343) claims that "if you compiled IP masquerading directly into the kernel, you do not have to do anything to configure it." (...referring to whether or not you have to load any IP Masquerading modules). This is clearly untrue as the ftp module, for example, still needs to be loaded as a module regardless of whether IP Masquerading was compiled directly into the kernel or not.

- The author says "you cannot use ping to test masquerading, because the ICMP...protocol doesn't have a source-port field." (p.346) Theoretically, true. Realistically, you can ping through IP masquerading till your heart's content. The fix has been around for some time now and is a standard feature of ipfwadm.

The problem with such mistakes is that they appear not a few times throughout the book, but many times. These are but a few examples.

Overall, the book proves to be a mere (albeit good) compliment to man pages, newsgroup postings, and howto's, not only because many of its explanations are incomplete and/or ambiguous, but because the reader suspects the need to confirm the validity of some of the information given. Moreover, and to the point of aggravation, the author repeatedly points the user to the man pages to get a complete understanding of how things work.

My recommendation: buy the book. It is extremely helpful and I recommend it. But beware. The recklessness with which things are laid out and the glaring mistakes resulting from this, warrant that the reader approach this book with a very cautious study.

--Daniel