For instance, a book on .NET 2 is up-to-date only while .NET 2 is up-to-date – once it has become an out of date language, then the book is out of date.

I fear that it is unfortunate that many people in the software development circles feel that many books are out of date – when they should be referring to more books that are in date. For example, early on in a programming career it is useful to read books on algorithms, design and optimization (particularly Big-O notation), but a book on “1001 ways you can use Version 2.1.342 Build 3 of HotProduct” is almost certainly useless, unless you actually need to know how to use that product – in which case the product’s website is more likely to be useful.

The most useful books are those that discuss trade offs between historical choices (whatever they are), why they were good at the time, and why have they proven to be a poor choice today. (For this reason, I find reading about early Unix, and early Windows design and implementation to be quite interesting, and likewise when it comes to programming paradigms.)