FHIR Chat · Removal of kind = Logical · fhir/infrastructure-wg

Stream: fhir/infrastructure-wg

Topic: Removal of kind = Logical

Grahame Grieve (Jul 02 2018 at 18:47):

in GF#17185, we agreed to remove kind = logical, and just have resources/types we define, or other things other groups define... but I'm stuck trying to make that real in the build tool. The problem is that we have a series of structures that we define that are neither types or resources:

MetadataResource
Definition, Event, Request
Data Element definitions for every element

Grahame Grieve (Jul 02 2018 at 18:48):

so I think we need to revisit this - I can't make the build / validator work without a way to mark these as not real types, and nor can I easily move them out of the FHIR namespace - which would be wrong because they are in the FHIR namespace

Grahame Grieve (Jul 02 2018 at 18:49):

And I'm stuck.... I haev to resolve this immediately or I can't build or do anything.

Ewout Kramer (Jul 03 2018 at 08:20):

These are quite exceptional - is this doable by having some kind of extension on the kind? I am hesitant of adding a code back to the valueset of kind, since I expect that at some point we'll have more insight in what these things are and have in common, and only then know how to handle them correctly. And then we would be stuck with an obsolete code in the kind valueset.

Grahame Grieve (Jul 03 2018 at 11:11):

well, I don't think that the concept that it is not a real resource is obselete. Exactly how it is not - that's the bit that's a moving target

Lloyd McKenzie (Jul 03 2018 at 12:39):

We'll be having more of them in the core spec over time, not less.

Ewout Kramer (Jul 03 2018 at 14:25):

We'll be having more of them in the core spec over time, not less.

I fully agree with both statements. Exactly how the resource is not a real resource (and thus would we be interested in having several categories of non-resources/non-types) is unclear, so I am afraid that finding a good term for the 'kind' of these things right now is hard.

Lloyd McKenzie (Jul 03 2018 at 14:27):

Not a resource = can't send it over the wire and not an abstract resource

Ewout Kramer (Jul 03 2018 at 14:44):

Lloyd, are you suggesting having 'not-a-resource' as a kind, or is this communicated using another mechanism?

Lloyd McKenzie (Jul 03 2018 at 15:06):

I'm defining what I think "logical model" means

Grahame Grieve (Jul 03 2018 at 20:06):

certainly that's what the tools need: this is neither a data type nor a resource, but something else. What exactly else is not important to the tools. But I can't do a build without making that differentiation, and our disposition was wrong to say that tings in the FHIR name space do not fall into that category

Ewout Kramer (Jul 04 2018 at 11:20):

"other" would truly suck. "pattern" is probably too limited as there may be other kinds of non-(resource/datatype). But I can see Grahame needs something now. I think we should find something that will not be in our way (normative compatibility wise) when we learn more about these distinctions.

"conceptual", "structure"?