The R28 issue, ID and altLabel

[BrianKingNOC]

Here is what @eric put in an email, so we can bring the discussion here to move forwards

Thanks for looking into this but NVS should not deflect this issue those who want to validate metadata – that’s what the NVS was designed to do. I see a couple possible solutions that NERC can implement:

1. Fix R28 by replacing IDs (CB0008,…) with the labels already in use by Argo (APMT,…). Since there are 1000’s of Argo metadata files coded this way, R28 should not be the only deviant.
   Or…
2. Add an NVS API allows users to pass the altLabel as an alternative to the term ID for validation. Then NVS will search the table for the label, and return the same data (HTML, JSON-LD, Turtle, or RDF/XML) as if it has been called with a Term identifier. This requires some server coding and a commitment to unique altLabels as well as term ID’s. It might look like:
   In my code, I would search using termID, if it is encoded fully qualified URL, e.g.,

“CONTROLLER_BOARD_TYPE_PRIMARY” : “SDN:R28::CB0008”

Validate with https://vocab.nerc.ac.uk/collection/R28/current/CB0008/

If it were an unqualified value from an existing float

            “CONTROLLER_BOARD_TYPE_PRIMARY” : “APMT”

Validate with something like: https://vocab.nerc.ac.uk/collection/R28/current/?altLabel=APMT
If query term (e.g., APMT) is not among the altLabel’s for R28, NVS would return HTTP code 404 “Not Found”

So, if NERC is not going to make tables backwards compatible (option 1) then they have responsibility to “fix it” with something like option2.

[BrianKingNOC]

At first I said that I thought the NVS team wanted to avoid option (1), because it doesn't reflect preferred practice. Since then I heard more feedback that if the community wanted to have the real-world labels in the ID, then that would probably be OK. So R28 could be made to look like all the others, and we can keep validating on ID.

I think R01 and R28 have the values that Argo uses in altLabel. The other tables have the value that argo uses in ID

As regards option 2: If the API can be made to validate on altLabel, then Violetta told me it would be easy to create and maintain a set of unique altLabels. In most of the tables, there is already an altLabel that matches the ID. That could be tidied up and maintained. So this would be a clean solution, and allow future tables to have the ID take the form CBxxxx. Existing tables could be changed to have CBxxxx type IDs as and when NVS is sure that it won't break any APIs.

[SBS-EREHM]

Note that altLable is just the final suffix of the termID "SDN:Rxx:SUFFIX".

The prefLabel is the human-readable label, e.g., "Argo meta-data".

[SBS-EREHM]

Update based on two SPARQL queries I got from NOC:

• The SPARQL query in the current Python validator checks the prefLabel.
• This updated SPARQL query checks the altLabel.

Actually, we have two cases:

1. Case where the prefLabel has been used in Argo Metadata instead of the Identifier. Actual example from an xxxx_meta.nc file:

platform-SBS-NAVS_EBR-1101 where

“DATA_TYPE” : "SDN:R01::Argo%20meta-data"

termURI : http://vocab.nerc.ac.uk/collection/R01/current/META/
Identifier: “SDN:R01::META”
prefLabel: “Argo%20meta-data”
altLabel: “META”

Checking the termURI or the altLabel would fail. Checking the prefLabel would succeed

2. Case where the altLabel has been used in Argo metadata instead of the identifier. Actual Example:

“CONTROLLER_BOARD_TYPE_PRIMARY” : "SDN:R28::N1"

termURI: https://vocab.nerc.ac.uk/collection/R28/current/CB0001/
Identifier: “SDN:R28::CB0001”
prefLabel: “N1”
altLabel: “N1”

Checking against termURI would fail. Checking against prefLabel or altLabel would succeed.

Summary: Since checking the prefLabel solved cases #1 and #2, I am using a SPARQL query that checks the prefLabel, not the altLabel.

So, that’s a fine kettle of fish. How do you all want to do handle these two cases? We could check both, I suppose.

[SBS-EREHM]

This may be moot given that PLATFORM/DATA_TYPE was removed. That said, I see no harm in checking both prefLabel and altLabel.

Please advise.