EDIT: My prerequisites are the classical set theory up to Goedel constructible sets, first-order logic up to incompleteness, and a (first) course in model theory.
My professor says that the Third Millennium edition (2003) is worse written than the first (1978), which he worships. I'd like to know your opinion about this: he says that the necessity to write an "incomplete" book (the subject has increased so much since 1978!) has forced the author to result faster and less clear in some passages. Is it a problem regarding only the added parts, or the changes in this later edition affect also the writing of old parts? In the first case I'd just buy the third edition and skip the new parts (which by the way I won't need in this course).
What about the second edition (1997)? Is it (substantially) equal to the first or does it have changes?
Of course I'm speaking from the mathematical (and didactical) point of view.
Thank you very much.
The first and second are the same. The third contains much new information. However a ${\bf Warning:}$ many of the new proofs in the third edition are incomplete referring at some essential point to a paper. If you then apply yourself to the paper it is very far from clear or easy to read. The exception to this is the material on pcf theory, which is a complete exposition due to Jech and which had been circulated for some time before the third edition as an easy entrance to pcf theory. (and is very good.) So I personally like to read the original and for a student I would recommend it. Of course you may also want to have the third to see some new developments without expectation of a full exposition.