Stroustrup Says C++ Education Needs To Improve
Stroustrup's book has passed the test of time. It's the only C++ book I own and I can say (without modesty) that I'm a fairly proficient C++ programmer. The book is the right mix of examples and is a very adequate reference.
I come from a scientific/engineering background and generally, when you truly want to understand and get the scope of something, going to the initial author or visionary helps a lot. Not only is their explanation simpler and more consistent than later development, it stays relevant as time goes. Especially when it comes to software engineering, one of the major problem to be solved is one of trying to simplify and generalize. So anything that isn't achieving that has probably failed.
I think C++ is simple enough and powerful enough for most needs. There are some quirks, but the language itself is sound and manageable. Most of the problems generally show up with libraries, they generally don't help to solve a problem (at all). While it may be different for other languages, it certainly isn't for C++ and once you remove libraries from your considerations, C++ becomes a very neat way of expressing solutions. It could be more concise, it could be even more generic, but I think a fair balance as been reached. Now if we could just have some half decent refactoring tools, that would be far more useful than this c++1x crap. Yes it's c++1x, I certainly remember how long it took compiler writers to get templates working right.