vector_base inheritance vs composition

Question

Short Question:

Is there a reason why C++ STL implementations use a vector_base struct/class (to handle resources and allocators) as a base class of std::vector rather than using composition?

Longer Version:

In my "quest" to improve my knowledge of C++ I've been trying to re-implement a Vector class, mostly std:: compliant. I think I've understood well enough why is sensible to use allocators and why you'd actually want all the memory handling in a separate class/struct (RAII and all that) but I fail to see why would we want std::vector to inherit from that class rather than having it as a private member. LLVM and gcc both use inheritance for example. I find, on the other hand, that constructors and assignments operators (especially the move-types) are much easier to handle using composition.

Am I only exposing my superficial knowledge of the language?

As an example I searched through some textbook and I found that Stroustroup's books have different versions of his "mock re-implementations" of the standard (just to confuse me more!) with his 2013 book "The C++ Programming language, 4th ed." using composition and his 2014 "Programming: Principles and Practice, 2nd ed." using inheritance!

Anyone can help me shed some light?

Answer 1

libc++ uses private inheritance and libstdc++ uses protected inheritance.

Protected and private inheritance is largely a syntactic variant of composition.