Esmvaltool: Start talking of datasets instead of models

[jvegasbsc]

There is one thing in the namelists that bothers me: we name all data inputs options as if they were always models, even the additional_models and reference_models that usually are observations or reanalysis. I think we should start talking of datasets instead, to avoid confusions.

I know this may seem really minor compared to all the issues we have to solve yet, but this is the kind of change that will become more difficult in later stages.

[mattiarighi]

Good point. Let's discuss this next week.

[bouweandela]

Having a good name for a single variable from a single model/observation would also help make the code much more readable. I'm not sure if dataset is such a great name though, because it's still pretty vague (and for me at least, confusion is possible with a netCDF4 Dataset).

[jvegasbsc]

I agree with you. In general, we should discuss this not only for the "model" ones but also for all the namelist options in general, as they will be more or less fixed after the alpha release

[mattiarighi]

Outcome of the discussion:

• we agree that models has to be changed for the reasons noted above
• several proposals were made, all of them with pros and cons
• datasets, despite the ambiguity it may have in some contexts (e.g. NetCDF, ESGF), is still the preferred solution and the most intuitive name for most users

@ESMValGroup/esmvaltool-coreteam