Semantic Web Health Care and Life Sciences Interest Group Teleconference

24 Jun 2015

See also: IRC log


David_Booth, EricP, Lloyd_McKenzie, MarcT, Paul_Courtney, Rob_Hausam, Tony_Mallia
David Booth


<trackbot> Date: 24 June 2015

FHIR element ordering

Comparison of approaches: https://goo.gl/DKcm0U



rdf: rest / rdf:rest / rdf:rest / rdf:first

<ericP> +1 +0 -1

Who prefers which option?

option 1

<Lloyd> +1 +0 -1

<Marc> +1 -1 -1

paul: +1 0 -1

<Tony> +1 -1 -1

rob: option 1? (abstain)

+1 +0 -1

AGREED: Option 1 for ordering

tony: what namespace to use for fhirRdf:index ?

marc: good to have one namespace

ISSUE: What namespace to use for fhirRdf:index ?

<trackbot> Created ISSUE-13 - What namespace to use for fhirrdf:index ?. Please complete additional details at <http://www.w3.org/2014/HCLS/track/issues/13/edit>.

lloyd: within repeating elements, the structure definition allows repetition, and round tripping must retain ordering.
... suggest fhirRdf:_index , because we might have the word "index" in several places.
... also calls out that this is a special beast.

AGREED: Start fhirRdf:index with 1

lloyd: everything else would have a class prefix on the local name, like fhir:Observation.code , whereas this would not have a class first.
... We could say this is a property of Element.
... there's a certain appeal to saying fhir:index

eric: when we start relating property names between classes, when two properties have the same meaning, how would we name them?

dbooth: My initial thought is that we wouldn't rename properties that are in common -- the ont would just indicate that they are the same.

AGREED: Use fhir:index (and change later if we need to)

Side-by-side example of two FHIR RDF approaches



lines 459 and 464-465

dbooth: on line 459, if you're using FHIR XML and you do an HTTP GET of http://record#PeterPatient , what are you supposed to receive? FHIR XML?

tony: you get back a patient resource, in FHIR XML or whatever format you requested

dbooth: on line 463, where did fhir:Reference.display come from? It isn't in the FHIR XML?

<ericP> <element> <path value="AllergyIntolerance.recorder"/> <min value="0"/> <max value="1"/> <type><code value="Reference"/><profile value="http://hl7.org/fhir/StructureDefinition/Practitioner"/></type> <type><code value="Reference"/><profile value="http://hl7.org/fhir/StructureDefinition/Patient"/></type>

<ericP> </element>

tony: I'll add that after line 459

<ericP> <element>

<ericP> <path value="AllergyIntolerance.recorder"/> <min value="0"/> <max value="1"/>

<ericP> <type><code value="Reference"/><profile value="http://hl7.org/fhir/StructureDefinition/Practitioner"/></type>

dbooth: When you GET http://record#PeterPatient , does the FHIR RDF that you get back identify the patient using that same URI? We need to nail that down.
... At the moment on line 465 it looks like you are assuming that the patient will NOT be identified using the same URI.

<ericP> http://piratepad.net/hcls

eric: What if you get back a bundle?
... How do we know if a bundle is permitted?

tony: Don't know. Haven't got to bundles yet.

eric: There's raging debate in the RDF world about whether to distinguish between a document URI and the thing it describes.
... The people who care about making a distinction do it in two ways: 1. frag ID (with media type). 2. Use a 303 redirect.

tony: Recommendation?

eric: we should be able to compose the RDF URI for the patient from the info in the FHIR XML

tony: why would the resource ID in the document be different from the document ID?

eric: When there is some logic in the object model, and that logic allows you to deref a URI, then we need to emulate those semantics in RDF.

tony: agreed

eric: If I get RDF for a patient problem, I think there's some glue in the protocol that says that if I do a GET on a URI, the FHIR data you get back is that thing at the top of the XML hierarchy.
... If we say that FHIR RDF is isomorphic with FHIR XML, then we need a way to identify the starting node, analogous to the top level node in XML.
... If I get some FHIR XML, with a reference, I need to be able to compose the URI for Peter Patient from it.
... That link needs to point to either something in the graph or something that's created by the conversion process.

dbooth: When <patient> has a <reference> to http://record#PeterPatient , the system that does a GET is expecting a Patient to be returned. That expectation is what allows the system to look for a Patient in the data that's returned.

eric: what if there are two patients in the data returned?
... The FHIR RDF will need to contain a pointer to its root element.
... I believe a resource ID can also be served from multiple URLs.
... With FHIR XML, when you get back FHIR XML you know what the root element is.
... Normally the URL at which you GET some RDF is not a URL in the returned document.

Paul: Suggest unpacking the GET problem in to a small example, as specific as possible and grounded in some content. Hard to follow.

eric: I'll implement it in the translator and show you.


Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.140 (CVS log)
$Date: 2015/06/24 16:39:29 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.140  of Date: 2014-11-06 18:16:30  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

No ScribeNick specified.  Guessing ScribeNick: dbooth
Inferring Scribes: dbooth
Present: David_Booth EricP Lloyd_McKenzie MarcT Paul_Courtney Rob_Hausam Tony_Mallia
Found Date: 24 Jun 2015
Guessing minutes URL: http://www.w3.org/2015/06/24-hcls-minutes.html
People with action items: 

[End of scribe.perl diagnostic output]