Can anybody explain to me why there is a difference between these two statements?
const A& a = A(); // correct
A& b = A(); // wrong
Non-const references must be initialised with l-values. If you could initialise them with temporaries, then what would the following do?
int& foo = 5; foo = 6; // ?!
const references have the special property that they extend the life of the referee, and since they are
const, there is no possibility that you'll try to modify something that doesn't sit in memory. For example:
const int& foo = 5; foo = 6; // not allowed, because foo is const.
Remember that references actually have to refer to something, not just temporary variables. For example, the following is valid:
int foo = 5; int& bar = foo; bar = 6; assert(foo == 6);