You can't derive a specific ordering of the Hamiltonian; the ordering of operators is a quantization ambiguity which has no classical counterpart. Usually you have p²/2m for the kinetic energy of a particle but there is no qm principle which tells you that (px)(p/x)/2m is wrong. Different ordering schemes may result in different physics and you have to use an additional, independent physical principle in orer to select the "correct" one.
In case of a curved manifold with a free particle moving on that manifold one reasonable idea is to use the Laplace-Beltrami operator as kinetic energy; this results in a unique operator ordering.
Perhaps Srednicki explains something like that ...