How should users define large amounts of dump variables?

[charlotte]

Testers: Met Office (Mark)
Type: Defining Inputs
Location: Files

Our dump files used for initialisation contain too many variables to define in the file forms. What approach should we use to defining dump variables?

Requires discussion.

[charlotte]

Len Shaffrey (HiGEM UREAD)
Would like to be able to bind files to input requirements, rather than only the variables contained within.

[bryan]

I can't progress this: as it says, it requires discussion. I don't really know what the problem is: either folk want the information to be captured, in which case they do it, or they don't, and they don't ...

[charlotte]

I guess the user may not know all the variables that the model requires from the dump file.   Maybe the exact input variables are not important (unlikey to be used as searched criteria) for baseline runs with a long spinup.

I think Len would like to be able to point to a file because that is what he does in the UMUI.  But this has no meaning for other people except for Len.  So now I'm back with Bryan.  Either the user enters the variables or they don't because it is of no value to other users to enter the file name without the variable information.

But Mark's issue is different, he wants a method of describing varibles that is somehow easier to handle than the method we used in beta1. 

Bryan, have you made any changes to file handling for beta4?

[charlotte]

We've changed the requirement on input information.  We'll only ask users to tell us about important initialisation fields as they see fit.

It is not necessary for users to tell us about ALL of the inputs required by their model!