The time ordering for the purpose of quantum mechanics is e.g. given by
$${\mathcal T} \left[A(x) B(y)\right] := \begin{matrix} A(x) B(y) & \textrm{ if } & x_0 > y_0 \\ \pm B(y)A(x) & \textrm{ if } & x_0 < y_0, \end{matrix}$$
where $x_0$ and $y_0$ denote the time-coordinates of the points x and y.
Now strictly speaking, the above definition works by pattern recognition - I can tell the operators $A(x)$ and $B(y)$ appart. I could either imagine one could write down the quantum mechanical theory by repacing all occurences of $\mathcal T \left[A(x) B(y)\right]$ by function notation $\mathcal T \left[A(x), B(y)\right]$ or that there is a computable procedure to make sense of $\mathcal T$ whenever I'm given an operator $F(x,y)$, i.e. $\mathcal T \left[F(x,y)\right]$.
Which is the case?