Reading yet another book which advocates
and getting really sick of it. Let’s state that: both are equally evil. The main advantage that is usually named when promoting get_b() is that one day you can change the implementation to do a little more without having to update the clients.
Once you change the semantics, you also must change the name – if your method suddenly starts doing a little extra, to be on the safe side you need to ensure that all users are fine with the new logic.
Of course one never can check all clients of a shipped library, but this only means that in a shipped library you’re better off not having public members (or getters) at all.