Theo Schlossnagle’s Scalable Internet Architectures is a book full of good ideas. Unfortunately, the writing is terrible and the intended audience doesn’t exist.
This is to say, there is nobody on earth that’s going to get value out of a whole bunch of example stub configuration files, when the very projects mentioned by Theo include such things and ample documentation.
Further, the writing is full of useless metadata, inconsistent typesetting, pointless digressions, ego stroking for the author, and pages of simple configuration examples that any half-competent system administrator or programmer will find tedious and pointless.
All of that said, I’m glad I read the book, and would recommend it, as I’ve yet to come across a better example for the genre.
What it comes down to is that Theo’s ideas on the difference between availability, load balancing, scalability (in both directions) and being cost-effective are excellent. If the book were written in the style of The Algorithm Design Manual I think it would be excellent. The author would describe a problem (DNS locality and high availability, say), and then describe the white paper solution (proprietary expensive hardware), the general terms of a better solution (virtual IP pools and anycast DNS), and some specific pointers to projects that may help (wack-a-mole). Or, hell, do the Design Patterns thing.
So, buy the book, read the ideas, ignore the examples, ignore the horrible writing, keep the specific implementations in the back of your mind, and have a good time of it. A better-written tome with the same ideas could be half as long and far more meaningful. 5/10.