Two different types of physics schemes

[tommbendall]

It would be really helpful if we could formalise a way to describe physics processes. There are two main paradigms:

1. we follow some procedure (possibly complicated) to evaluate a source/sink field, because there is no neat expression for the physics process;
2. we can write down a neat expression for the physics process, so should be able to solve it with any time discretisation that we like.

All schemes are currently written in form 1, but a handful have hacky methods to be like 2. However we may also find schemes that mix the two situations -- e.g. a wind forcing of the form F = -C*u, where C is a coefficient that has a complicated dependency on other prognostic fields.

Hopefully this discussion can help us find a solution.