Compilers - Principles, Techniques and Tools [Hardcover]

Compilers are a very technical subject, and it stands to reason that even with the best books, youre gonna have to read certain sections more than once to grasp the content.

Get this book in paperback if you can. Its much easier to flip through, which youll find yourself doing on a regular basis.

The former kind is more suited for a course on theoretical aspects that lay the foundation of compiler construction. DFAs, NFAs and Regular expression along with relations and equivalence between the them; FSAs minimizations; grammars and Push-down FSAs in details, ambiguities and and how to cope with them; and so on.
This is what I mean for "theoretical aspects". And these topics are covered in great details in this book. Almost the same details they (the authors) placed on writing a more specific book as "Introduction to Automata Theory ...".
Same situation applies to principles on more application- oriented topics. Take the example of LR parsing. You can face the topic from a more theoretical side, dealing with details on bottom up parsing (still, it implies an in-depth knowledge of grammars theory), handles and (viable) prefixes, SLR or canonical LR or LALR parsers and techniques for the relative tables construction by hands (and for this, add a detailed and solid knowledge of Push-down FSAs along with grammars). By hands, at least, if principles are what matter in your course.
If you expect to find these topics (with this depth) in a book of the other kind, you might get mislead. As I did when I still had not clear this distinction, before I took the course.

The latter kind of books is more suited for a more pragmatic course. One where real, "modern" compilers are at hands, and probably written as homeworks. In this case, time being always limited in a university course, one (the instructor) will likely have to give up with those theoretical aspects (or just mention them) and focus on coding techniques and modern compiler studying. But ... perhaps, for these purposes books like Grune's "Modern Compiler Design", or Pittman's "Art of Compiler Design, The: Theory and Practice" or, at some degrees, Muchnick's "Advanced Compiler Design and Implementation" will be more suited.
Back to the LR parsing example, more pragmatic compilers design courses will (for time constraints) just have a glance on principles and spend a considerable time learning YACC. To do both things you would have to take a course on YACC alone (it requires time to exploit all of its advanced features, you can be sure of this).

All this said, once again: which is the best book ? The one that best fits your needs. And in fact, my needs were those of my course, which was completely centered on theoretical aspects. And for this course, the Dragon book (as it is better known since its cover) proved to be perfect, definitely no matter it was published on 1986: FSAs and grammars and their theory is (for all practical purposes) still the same since even before then.