See also: IRC log
<trackbot> Date: 01 April 2015
<hsolbrig> The dial in is being crabby
<ericP> crabby or just reluctant
<ericP> ?
<hsolbrig> Crabby in "This is the Zakim conference bridge... followed by "ring ring ring" followed by a hang up"
<hsolbrig> reluctant
<hsolbrig> Not sure what I can do when Zakim refuses to talk to me
harold, try the code right after the first message
scribe: before it tells you to enter it.
<Claude> https://global.gotomeeting.com/join/157514853
lloyd: In XML, sibling ordering
matters, but not in JSON, so not in RDF either.
... Where we cannot redetermine order from the metadata is in
repeating elements, such as 5 names.
... Order must be maintained even if the spec doesn't
explicitly say that it is semantically important.
... Dig sig will need ordering too.
eric: that's doable.
<twamarc> sorry keeping being disconnected ---don't know why!
lloyd: If the cardinality is more than one, then ordering must be retained.
<Lloyd> http://hl7-fhir.github.io/observation.html
<ericP> <<a title="Who was responsible for asserting the observed value as "true"." class="dict" href="observation-definitions.html#Observation.performer"><b>performer</b></a>><span style="color: Gray"><!--</span> <span style="color: brown"><b>0..*</b></span> <span style="color: darkgreen"><a href="references.html#Resource">Resource</a>(<a href="practitioner.html#Practitioner">Practitioner</a>|<a href="device.html#Device">Device</a>|<a href="organizat[CUT]
<ericP> from observation.profile.xml
ISSUE: What RDF mechanism to use for lists/sequences?
<trackbot> Created ISSUE-5 - What rdf mechanism to use for lists/sequences?. Please complete additional details at <http://www.w3.org/2014/HCLS/track/issues/5/edit>.
<ericP> <b>0..*</b>
<Lloyd> http://hl7-fhir.github.io/observation.profile.xml.html
<max value="*"/>
<ericP> //element[path@value="Observation.performer"]/definition/max
<ericP> /Profile/structure/element[path@value="Observation.performer"]/definition/max
lloyd: Any max other than 1
requires ordering to be preserved.
... Ecore in Eclipse is useful for code generation
eric: Any technique for property equivalence?
lloyd: FHIR doesn't capture property equivalence
<ericP> Lloyd: W5 column asserts the primary object and the timing of a resource
<ericP> ... more of a subproperty than an equiv
<ericP> Lloyd: e.g. {ValueSet,StructureDefinition,ValueSetDefinition}.publisher are logically equiv
<Zakim> dbooth, you wanted to suggest distinguishing between basic mapping needed for round tripping versus icing helpful for inference and other use
<ericP> Lloyd: i don't see equiv property as part of initial pass
<ericP> Lloyd: i expect that round tripping the instances will involve minimal icing
eric: let's look at tony's work
<ericP> https://global.gotomeeting.com/join/157514853
tony: looked at common properties
across FHIR resources
... counted about 230 of them.
<ericP> Lloyd: these [equiv?] aren't safe
lloyd: None of these are
safe
... Every place you see the property 'code' or something else,
it could mean something completely different.
... There's no way from the FHIR metadata to determine whether
they really are the same property (semantically)
... Even W5 does not indicate sameness.
tony: AllergyIntolerance has a recordedDate
lloyd: But it's really AllergyIntolerance.recordedDate -- fully qualified.
tony: Could map all of the fully
qualified FHIR names to RDF URI
... or could may non-fully-qualified names to RDF URIs and then
distinguish them by the context of their usage.
eric: The sweet point that ontologists try to get is to have common properties that have restrictions on their use.
lloyd: Yes, but not restrictions
on meaning.
... fhir:date could mean: 1. time last changed 2. time the
action was performed 3. time sent 4. time the entity came into
being.
... These are different meanings.
claude: I don't like the naming of attributes as types -- not informative. You don't know what it is.
lloyd: We have elements called things like 'date'. We don't decorate names when the context makes the meanign clear.
<ericP> Claude: we have to autogenerate this mapping so we can't rely on knowledge from outside the definition process, we won't be able to do this by machine
<ericP> Lloyd: "code" isn't safe; needs to be "AdverseReaction.code"
dbooth: this is grouping them by their spelling, rather than by their semantics
<Claude> I would argue that if you see ValueSet.date, this is not informative from context. I would need to read the definition to understand its meaning
<twamarc> having always umbrella property on top reject obviously the idea of having generic properties. isn't it?
FHIR has no generic properties (or at least almost none)
<Zakim> dbooth, you wanted to say this is exactly one of the problems that we faced in the JSON-LD approach
+1 to fully qualified names
AGREED: Use fully qualified FHIR names for RDF properties
eric: Should also show example of the benefit of non-fully qualified names to motivate FHIR spec to add info on equivalences between properties
lloyd: Asserting equivalences can be either with short (common) names or with other (explicit) equivalence relationships
oops! out of time!
lloyd: Need to address how to
deal with identifiers and codes.
... Might be useful to go through element definitions and see
what is needed for RDF.
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 Default Present: DBooth, ericP, Tony, lloyd, Marc_Twagirumukiza, rhausam, [IPcaller], claude, harold Present: DBooth ericP Tony lloyd Marc_Twagirumukiza rhausam claude harold Found Date: 01 Apr 2015 Guessing minutes URL: http://www.w3.org/2015/04/01-hcls-minutes.html People with action items:[End of scribe.perl diagnostic output]