Oops, I downvoted you when I meant to upvote. There is probably another incarnation or three before the one you linked, too.

I'd recommend Appel's "Modern Compiler Implementation in ML" over the dragon book.

I noticed there are a few versions, including for Java and C. Any reason why the ML version is better (not against ML; just don't know it well enough (at all actually)).

A very large part of compiler operations involve analyzing and transforming tree/graph data structures (once you've parsed to an AST), so you want a language with good native support for those structures. Pattern matching and garbage collection help immensely, as do ML's inferred static types. Working with complex data structures is really where ML shines.

It's not just ML, though - using a language where you can work directly with native tree structures (ML and Haskell's type constructors, Lisp sexps, Prolog functors, Lua tables, or even JSON) means that you can postpone learning syntax/parsing and start with the language's semantics. You can save the syntax details for when you have the language design worked out. Compiler books based on C (such as the dragon book) spend such a long time on parsing upfront because they don't have much choice.

Also, Andrew Appel works on SML/NJ, and the C and Java books are translated from the ML version. I've looked at the C version (which uses lex and yacc instead of ml-lex and ml-yacc, mallocs but doesn't free until later in the book, etc.), and while it's still a good compiler overview book, I'd strongly suggest getting the ML version instead. The ML version uses Standard ML (SML), but I've only used OCaml and didn't have trouble following it and making the minor adaptations along the way. (I haven't looked at the Java version. Whatever.)

I think that ML has some features (such as pattern matching) that makes it a better fit for writing a compiler than Java or C.