A specific advantage of this book is that it is one of the few that dedicate a whole chapter on the connection between Bayesian methods using Gaussian Processes and Reproducing Kernel Hilbert Spaces. Even if this connection is a posteriori pretty obvious, it is nice to have it broken down clearly into small understandable pieces.
Otherwise, all the explanations concerning Gaussian Processes themselves for regression and classification are very clear and make this book a very worthwhile read. I would recommend also reading other books focusing more on Reproducing Kernel Hilbert Spaces in order to have a complete picture of these methods (e.g. "Learning with Kernels" by Scholkopf and Smola or for an even broader picture "Generalized Additive Models" by Hastie and Tibshirani).
Finally, since GP and RKHS for classification are still evolving subjects, it is probably a good idea to keep reading more material on them after finishing this book.