I only believe in books which teach general computer science material, so if I'd have to pick books, I'd pick
- D. Knuth's work on algorithms.
- Code Complete, McConnell
- The original GoF book perhaps.
- Yourdon's work on information analysis.
- and because I'm a fan, T.A. Halpin's / Nijssen's books on information analysis and modelling. (around NIAM/ ORM mostly)
But I've been away from the curriculum's of current universities for a long time, so I don't know what the books of today are. What I'd like to see more as well is that people are made aware of CS research papers: that they search in Google scholar for papers about a subject and learn how to read science papers and how to filter them to get to the real gems (e.g., check abstracts, learn how they're build up and why the references are very helpful and often lead to a greater set of good papers to read)
For example, I have the arXiv RSS feed on CS papers in my google reader:
http://export.arxiv.org/rss/cs
It gives a new fresh list of papers every 2-3 days (10-20 per refresh). Often they're very detailed, but sometimes true gems are in there, and what's more: leading to papers which are also worth reading through the references, papers which are often much older.