Agreed. I'm taking a class right now with the dragon book (taught by Aho, actually - one of the authors of the book), and it's easily the best and most important CS class I've taken[1]. The main project for the class is to design a language and write a working compiler/interpreter for it - in all the years he's taught the class, apparently no group has failed to do this. The really tricky part is making a language that other students in your class will want to use, though, and that goes beyond the 'simple' issue of compiler structure!
I half-wish that there would be a way to teach a compilers class as the second class in a CS track (ie, right after the intro class), but the problem is that it just requires too much familiarity with the details in order to impose upon most beginner/intermediate students, even if a few really interested ones could potentially handle it.
It might be easier if you started off with Scheme and then went from there (a la SICP), but even then, you'd miss a lot of the more fundamental concepts assumed in the Dragon book that really just need a lot of time to be digested.
[1] Alright, there's probably a three-way tie for 1st, but it still counts. My education would not have been complete without it; I highly recommend the Dragon book to anybody who has the time and interest.
My compilers class was a nightmare (and we used the Dragon book). I had hoped a compilers course would actually teach about the full process by which a text file becomes executable 1s and 0s. Instead it was a months-long torture of parsing text. Regex to NFA, NFA to regex, regex to DFA, NFA to DFA, DFA minimization, LL(1), LR(0), SLR(1), LR(1), LALR(1), along with memorizing all the rules governing which grammars are amenable to which parser, and working through all of the preceding by hand. And then finally about a week of teach-yourself-LLVM.
I don't know if "compilers are a solved problem", but I'm pretty damned sure that the only people who need to know that much about text parsing know who they are, and would be able to reference it when needed. As an undergraduate course, it was utterly worthless.
I interviewed Professor Aho a few years back; I'm a little jealous I was never fortunate enough to take that class. It's likely to be one of the most practical classes in all of computer science anywhere.
(I won't give away the punchline for the class. I don't know how far along you are on the project.)
I took that class from the other instructor, who had us use OCaml instead of Java (as I presume Aho still uses.) The Dragon book was interesting, but, for the most part, we were able to get on from the in-class slides.
I don't think there's anything wrong with having Discrete Math, Data Structures or Computer Science Theory as pre-requisites for PLT or a compiler class. However, I doubt the necessity of OO Programming & Design and (so-called) Advanced Programming.
I half-wish that there would be a way to teach a compilers class as the second class in a CS track (ie, right after the intro class), but the problem is that it just requires too much familiarity with the details in order to impose upon most beginner/intermediate students, even if a few really interested ones could potentially handle it.
It might be easier if you started off with Scheme and then went from there (a la SICP), but even then, you'd miss a lot of the more fundamental concepts assumed in the Dragon book that really just need a lot of time to be digested.
[1] Alright, there's probably a three-way tie for 1st, but it still counts. My education would not have been complete without it; I highly recommend the Dragon book to anybody who has the time and interest.