Today I'm getting back to reading SICP (Structure and Interpretation of Computer Programs) and Halliday and Resnick's classic Physics text.  These are sort of like The Odyssey for CompSci and Physics in a sense, not necessarily because they are widely beloved--although I get the sense that they are at least appreciated by most academics in the relevant field.  They're classics in the sense that almost everyone in the relevant field has read them.

I like reading classics because

1. Something must have made them a classic.  Especially the old ones that became popular before the textbook industry turned into big business with lots of corruption in university departments, gain a lot of prestige on this account.  Stewart's Calculus is fine I guess but it's not nearly as respected as it is widely used--and I suspect that some amount of kickbacks are responsible. 
2. They get everyone on the same page.  In order to meaningfully say that you know programming, it suffices to read a book (and do the exercises, and similar projects) like SICP.  Other books also suffice, like Knuth's monster, but it's nice that we're all on basically the same page about SICP.

I'd like to develop a running list of classics.  I think I'll go do that now and make a page about it.