Lagrangian is function of generalized co-ordinates, generalized velocities and time:
$$L=L(q,\dot{q},t)$$
Why the specific form $L=T-U$ is used as a definition of Lagrangian function? Here as usual $T,U$ are kinetic and potential energies respectively.
Why $L = \dfrac 1 2 m \dot{q}^2 + U(q)$ is not a Lagrangian function, even though this $L$ also is a function of $q,\dot q$?
Where did Lagrange write $L = T-U$ in his famous book "Mécanique analytique"? Or what justification in support of this definition, did he give there?