[saucetenuto]

I don't agree. The _signature_ should indicate whether a function has access to a class's internals or not, but the caller shouldn't have to care. This is already true in C++ today: when we write f(x), f might be a normal free function or it might be a friend of x.

Viewed from that angle, this proposal makes C++ more consistent, not less: it breaks a needless dependency between syntax and semantics.