Yes, it certainly would make a good post. But it is incomplete. The distinction between exogenous and external is fairly subtle, with exogenous indicating that the behavior originated from outside but now is part of the system. A GHG such as man-introduced CO2 is an exogenous factor. It was previously completely externalized, having been buried for eons underground as hydrocarbons, not having an effect on the climate. Once extracted, the hydrocarbons are combusted, which then turn into the exogenous CO2 and which in turn has significant impacts on the earth’s climate.
Just as critical is the idea of a compliant interface. That is an interface to an external entity whose characteristics change with contact. This is best illustrated by considering a vehicle in contact with a surface. A non-compliant interface would be if the vehicle interfaces with concrete. The concrete may budge slightly but can be considered static. A compliant interface would be a vehicle interacting with sand or mud or water. The model of the compliant surface becomes much more difficult to characterize simply.
In terms of modeling temperature, a hot or cold body is automatically compliant with an external thermal reservoir. On contact, two bodies at different temperatures will redistribute the heat according to laws of entropy. This is a very compliant interface that can be simplified only if the thermal heat capacity of the external object is large. This is significant in terms of OHC modeling, but simplifications are indeed possible since the ocean has an enormous heat capacity.
I have a more complete analysis under the Stochastic Analysis tab here http://contextearth.com/, and other documents as well.
This particular site is devoted to context modeling. A context model is a recently coined term to describe models of external interfaces that can be used to design systems that are used in a specific context — for example, an environmental context model of water is used to design a sea-worthy boat. It is a pedantic definition, but pedantic is good in this case as anything to enable separation of concerns is good. Separating the design & component models from context models fosters reuse.