I (finally) just finished Lee Smolin's The Trouble With Physics. There have been many reviews and lots of back-and-forth on the scientific blogosphere regarding its main theme: that String Theory (or more accurately String Theories) is a dead end unless the establishment realizes that more encouragement and backing needs to be given to those promising grad students who would rather explore quantum gravity and other non-mainstream topics.
Most of Smolin's excellent book in fact deals with just how strings came to dominate and suffocate physics over the past 25 years. I must say what struck me was the fact that current string candidates are mainly background dependent: meaning, they do not run with Einstein's most provocative idea, and one of the greatest achievements of general relativity--which is that it is background independent. Space and time are dynamic in Einstein's universe; indeed they emerge from his theory as important agents (if you will) in how it all works.
But all the currently mainstream string theories are background dependent, meaning they have to assume at the outset a static space in the background against which strings work. This to me certainly sounds like a step back...but no one in the past quarter century has figured out how to come up with a background-independent theory. The book is very well written, and in spite of the thesis--not at all obnoxious or arrogant. This too was a pleasant surprise, as I must confess, I couldn't get through Smolin's earlier book, Life of the Cosmos, which I found wooden in style.