Meeting minutes
Issues 77 and 93
https://
eric: On 93, option 2c, we currently have: fhir:origValueBoolean true
… Propose adding origValueBoolean also to #77 for regular extensions.
… Or just fhir:value
… In RDF we typically put the datatype in the value itself, instead of making a separate property name for each datatype.
… e.g., in RDF 'fhir:valueDecimal 0.75' is redundant, because 0.75 is already xsd:decimal
eric: could call it "extendedValue"
david: Or "originalValue"
eric: We cannot call it fhir:value because that's already used.
Action: Eric to write up proposed resolution for #93 and #77
Issue 76: How should list ordering be preserved in R5?
https://
jim: On the last call, people seemed inclined to stay with the R4 approach
eric: I'm slightly inclined to change to RDF lists now.
jim: Difficult to use the ordering in SPARQL
eric: True. But rarely care about the order. Only care when validating slicing.
… e.g., first entry in an address is a home address.
david: Do we have more usage data, on the impact this has on SPARQL queries?
eric: None of our NHS queries have cared about the order.
brad: I've never had to care about the order either.
james: Patient name, and patient telecon has the order preserved.
… But I haven't had experience needing to care about the order.
… I'm more working on generating the FHIR data.
brad: If a patient has multiple names, we only care about all the possibilities, we don't care about the order.
james: Might be important with labs or medications.
brad: Everything comes through with a datetime on labs, so we use that.
… People care about trending ... "give me the 5 most recent".
eric: People care in a slicing example, but shex takes care of that.
… JSON-LD 1.1 supports RDF collections now, so that would allow more of the conversion processing to be done using JSON-LD 1.1 rather than doing it in the preprocessing.
jim: RDF list is nice and concise. Leaning toward that.
Action: David to write up option 4, and new option 3 to be a regular RDF list
Eric: suggest using as example, a series of codings in a codeableConcept, or a series of addresses.
Action: David to change to a better example also.
david: Is there enough benefit to make this change to R5?
eric: We can make this change in our code. Now is the time to do it.
Issue 69: Shorten FHIR property names or use superproperties?
https://
eric: Want to remove the datatype from the property name: instead of fhir:valueBoolean it would be fhir:value .
… Instead of fhir:valueCodeableConcept it would be fhir:value [ a fihr:CodeableConcept ; ... ] for complex types.
jim: https://
eric: OWL DL won't allow classification based on a property value that is a literal.
ADJOURNED