I just found out that the original version of Nyquist-Shannon sampling theorem differs from the modern version..
In the original version, it states that the bandlimited signal $x(t)$ can be constructed by the sampling rate of $2B$ when $B$ is the highest frequency content of the signal.
However, the modern version states that the signal can be constructed by the sampling rate of $2B$ when the signal's highest frequency content does not exceed or equal to $B$.
So, is the original version wrong? Or are there any reasons why these two came to differ?