terms of the `processing level` draft vocabulary

[BaptisteCecconi]

The terms proposed in the processing level draft vocabulary are full text terms (e.g.: obscore-raw), whereas the ObsCore specification uses numbers from 0 to 4. Hence we can't will not be able to reference the vocabulary in ObsCore.

[molinaro-m]

Unless ObsCore is updated and a mapping is provided.
True that it won't be a painless transition changing an interface element from being a number to become a string.

My opinion is: have the vocabulary in place and let the ObsCore document pick it up in the best possible way, that might initially include a mapping inside the doc itself to say that, e.g. , 0 equals "obscore-raw" and, later on, simply use it (breaking change).

[msdemlei]

On Tue, Jul 08, 2025 at 01:42:16AM -0700, Marco Molinaro wrote: molinaro-m left a comment (ivoa-std/Vocabularies#40) Unless ObsCore is updated and a mapping is provided. True that it won't be a painless transition changing an interface element from being a number to become a string.

Hm... I don't think I'd like to go there. In principle, it's a minor thing, and I don't think many people constrain that right now, so I doubt we'd be breaking many *queries*. But as soon as some services have strings in a column and others integers, vstacking (or running a UNION with uploads or the like) query results becomes a pain (or impossible).

My opinion is: have the vocabulary in place and let the ObsCore document pick it up in the best possible way, that might initially include a mapping inside the doc itself to say that, e.g. , 0 equals "obscore-raw" and, later on, simply use it (breaking change).

Well, we *could* use '0'...'4' as labels and then appeal to automatic type coercion towards Obscore. That doesn't feel quite right, though, in particular because I would have liked to represent the processing_level table from EPN-TAP2 (p. 13) in that vocabulary, and there's just one '1' we can define in SKOS; so, EPN-TAP2, CODMAC, PSA, NASA and ObsTAP will clash. I think I'd already be happy if we got the EPN-TAP2 table represented in the vocabulary somehow. Obscore would then say "for is historical reasons, we do not use terms from the processing-level vocabulary but instead an integer. The integer values corresponding to vocabulary concepts are in paretheses at the start of each pertinent description. Pretty? No. Sure, we *could* define a property for defining the obscore level belonging to a concept, and even define the range of that property (i.e., the type of its objects) as integers. That would be clean. But I can't see a scenario where that would provide an actual benefit over what we already have.