C++ language despite of the powerful mechanisms supported is not a language for the faint hearted. Two forces drive its peculiar concept of friendliness (it is not unfriendly, just very selective) the backward compatibility with C and the effort to not getting in the way to performances. This book, written by the language father, presents and analyzes the language history and the design decisions. And, given the writer, the perspective you get reading the book is very interesting and more than once helps to shred some lights in the dark corners of the language.
The history is very interesting since it details how the language genesis and marketing went from the AT&T labs to the academy and industry.
C++ design principles are presented and the most notable is that of ease of teach-ability. Several time proposed/existing features had been modified or dropped entirely because they were not easy to teach.
Another very interesting principle is the “you don’t pay what you don’t use”, meaning that features added to the C language in order to define the C++ language were designed so that the programmer would not incur in any penalties if not using them. That’s why if a class has no virtual method, then the pointer to the virtual methods table is not included, saving the pointer space from the class instance memory footprint.
Aside from answering to many questions, the book opens up a bunch of new ones. For example, the very first implementation of C++ has been developed practically around a threading library. Now more than 30 years later, in a world with an increasing presence of multi-core machines, the C++ standard still lacks of a multithreading / multiprocessing facility.
Also Stroustrup asserts more than once that a Garbage Collection way of managing memory could be add by a specific implementation. But fails to explain how this non-deterministic way of terminating dynamic memory life could deal with the deterministic needs of destructors. Likely I’m just to dumb to figure out myself.
The big miss I found in the book had been a comparison with Java language. Basically one of the great contenders for the title of most widely used programming language. Java, on its side, has some interesting approach to language design that conflicts with those of C++ (e.g. the C compatibility issue). Therefore it would have been nice listen from Bjarne voice his thoughts about. In his defense it has to be noted that by the date of this book hit the streets, Java hype had just been started.
Last complain about the book is the lack of conclusions. The book seems cut a couple of chapters before the real end. Aside from stylistic point of view, some words about the future evolution and perspective would have been at their place at the end of the book.