Standardization of interfaces

[HendrikTolman-NOAA]

If we decide to support multiple options for solvers and source terms, standardization of subroutine interfaces will result in a code with a simpler and more maintainable overall architecture, but may result in sub-optimal optimization for some options. Do we want to make such standardized interfaces a core part of the architecture?

[JessicaMeixner-NOAA]

Some level of standardization will be helpful for being able to switch between source term and other auxiliary options. For example, if you are calculating a tail parameter that is not source term specific (FLD1,FLD2 in WW3) then there are assumptions that certain variables are calculated and if you switch between source terms without that, things break but not always in obvious ways.

In examining how to remove switches from WW3, having standard input/output to routines would have been extremely helpful -- however if memory is not all stored in modules and we don't have switches from the start, this might not be a requirement.