I've just counted, it's the fifth book I've read on C# this year (not to mention a dozen books on .NET) and this is by far the best. It's not just very good technically, useful and enjoyable to read, it's flawless. In fact, I didn't find as single typo, a single sentence that was slightly misleading or incomplete which, in my experience, is extremely rare for a first edition. On top of being a highly competent developer, Jon is clearly a gifted writer and a born teacher. Every term is always used appropriately and in the right context, every example is spot on and contains the least amount of code that shows the full extent of the feature... this is a rare treat.
So, forget about reading the C# Specification. If you want to know all the useful stuff without going through pages and pages of boring and tedious text, read `C# in depth' instead! (I've made the mistake of doing it the other way round).
To sum up, if you are a good C# developer and you want to become a very good C# developer, there are 3 books you can't afford not to read: Framework Design Guidelines by Cwalina/Abrams, CLR via C# by Jeff Richter and this one!

I bought this book because Eric Lippert mentioned it on his blog (he is one of the reviewers and wrote the foreword) and I'm a big fan of Eric's work. I guess reviewing a book is not the same as writing it! I started with a deep interest in both topics (C# 3.0 and design patterns) as well as (too) high expectations and I got fairly quickly disappointed by the lack of innovative ideas in the proposed implementations and the somewhat superficial use of C# 3.0.
The code lacks consistency as if it was written by different people with different level of expertise, the layout (of the code) is far from perfect with a few irritating typos. Sometimes the lengthy explanations about the implementation cover trivial matters while unnecessarily confusing bits are just taken for granted.
There is even one suggestion that made me fall from my chair, page 170 in the Chain of Responsibility and Command pattern: storing information in the numerical value of enums! What a convoluted idea for no practical benefit (apart maybe saving 4 bytes of memory) and a potential nightmare for maintenance.

Overall, all the ingredients for a good book, but you might have to wait for the second edition (if it ever gets that far) to make it worth it.