Standard thing: older textbooks are generally better. They tend to get dumbed down in new revisions, and modified to fit stupid standards about class syllabus.

This reminds me that I actually bought the first edition of the GC handbook on purpose, because it had more C++ material, which was relevant to my project.

The second edition cut a bunch of C++ material out.

And yeah I don't think I missed anything -- I think it even had better diagrams.

So yeah I believe that general rule! I didn't realize that at the time I bought the Dragon Book.

Pretty sad how you can lose knowledge over time!

---

The books also respond to fashion -- e.g. Java was a popular teaching language, so they had to add more Java. I think that is reasonable in some ways but can backfire in the long term.

Although I guess I am not too excited about MIX assembly in TAOCP and so forth, so it's a hard problem

Obviously sometimes the updates are fixing real critiques of earlier editions. Those are good. But many times it is “fixing” critiques along the line of “cut out advanced material because it is too much to cover in 1 semester.” Makes sense in an academic context, but not for professionals like you and me that are looking for that kind of stuff.

Books also tend to become more sparse on the page, and with more pointless “illustrative” stock photos that add nothing and take up page real estate (leading to more advanced topics being dropped to reclaim space), but make the book seem more approachable to undergraduates.

Generally applies more to physics, chemistry, and math books than computer science, but it is a general phenomenon.