@kwilcox and I have been talking about what additional baseline tests could be conducted, in an attempt to generate meaningful statistics from available endpoints. I came up with two, and maybe as a group we could come up with more. Any feedback, additions, or subtractions before aiming Kyle and pulling any triggers would be appreciated.
"Split Number of Service Results By Variable" by CSW source, e.g., as stacked bar charts with colors representing CSWs
The CoreVariable_Strings notebook shows number of services per variable, and number of services per CSW, but it's hard to tell which variables are held by which CSWs. You can start to see it in the Services per CSW server output but currently we're only showing NODC and NGDC. Plotting them all might help determine if there's a case where (for example) sea_surface_height is all from one CSW and sea_surface_height_above_geoid is from another.
Something getting at consistency (or lack thereof) of variable names
Variable names are key to data discovery. But anecdotally it seems like the discovery process can require searching for a number of variable names, some CF-compliant and some not. Can we develop a test for how many variable names across all the CSWs are CF-compliant and how many are not? If we keep track of which CSWs the names are coming from, we might see one or two standout CSWs that are doing particularly well or particularly ... not well. The results can start to put numbers on how easy discovery of data actually is, from the standpoint of a researcher with just a list of CF variable names.
@kwilcox and I have been talking about what additional baseline tests could be conducted, in an attempt to generate meaningful statistics from available endpoints. I came up with two, and maybe as a group we could come up with more. Any feedback, additions, or subtractions before aiming Kyle and pulling any triggers would be appreciated.
Anyone else have anything we might add?