I might be fast-forwarding a bit here, but this article paints a rather grim picture of the XML book business, an in my opinion many conclusions extend to other types of books on modern/complex/online technologies.

The article cites O'Reilly editor Simon St. Laurent, who shares his thoughts on the xml-dev list and later tries to understand what's happening

Clearly, for many people a simple subset of XML is enough. This might also be true for other similar technologies, and on the other end of the skills range you'll find people who are good enough at learning to not need how-to or best practices books. It's not surprising that API books are going down as well.

For me the highlight of the article is this conclusion:

Thus, doing something to lower the cost of producing a new book -- or finding a new genre of publication which is (a) cheaper and quicker to produce than a book; (b) more substantial than a long web article; (c) targeted at these interesting niches; and (d) not insanely expensive -- might help the bottom line.

I've been talking about this recently with Ceki Gülcü who is happily self-publishing his book on log4j. His experience makes me think that we'll see more and more authors going the independent way, in order to keep costs very low and make it economically possible to publish specialized books for much smaller audiences than in the past few years.