See also: IRC log
<trackbot> Date: 28 July 2015
<dbooth> Scribe: rhausam
<dbooth> july 14: http://wiki.hl7.org/index.php?title=ITS_RDF_Concall_Minutes_20150714
<dbooth> july 21: http://wiki.hl7.org/index.php?title=ITS_RDF_Concall_Minutes_20150721
dbooth: action item for Eric to write up about terminilogies
<dbooth> APPROVED: minutes of july 14 and 21
<dbooth> paul: no news
action item for Paul to write up PSS
<trackbot> Error finding 'item'. You can review and register nicknames at <http://www.w3.org/2014/HCLS/track/users>.
eric: won't be available for call tomorrow
<dbooth> david: Use ConceptBase (initially) as the superclass?
david: motion to use ConceptBase as superclass (a.k.a. "CodeableThingy")
<dbooth> i.e., what we called CodeableThingy would be ConceptBase
<dbooth> ... CodingThing is now CodingBase
<dbooth> ... and CodeThingy is now CodeBase
tony: have made all of the changes
david: motion to use those names, second tony
<dbooth> AGREED: use those name
paul: motion approved - unanimous
david: discussion of XSD decimal
precision
... the XSD decimal data type for expressing decimal number -
in the definition does not distinguish between 2.0 and
2.000
... in medical field number if digits typically implies
precision
... in FHIR, data type will be xsd:deciman, but note will
indicate that digits will be retained to and imply
precision
... how to represent in RDF?
<dbooth> xsd:precisiondecimal
david: several options available
- one option is xsd:precisiondecimal
... that data type also adds 3 special values - pos infinity,
neg infinity and "not a number" - those values will not occur
in FHIR, if using xsd_decimal underneath
... could define another data type
lloyd: can't do it for XML
representation
... not sure how tooling will handle it
eric: RDF can deal with alien data types
tony: can add xsd:precisiondecimal to tooling
david: if use precisiondecimal will not lose trailing digits
marc: how do reasoners use this?
tony: for any restriction, reasoner will figure it out
eric: OWL can't do < or >
david: if we do use xsd:precisiondecimal, would we lose anything?
lloyd: will the tools at least recognize that it's a decimal?
eric: SPARQL won't recognize
it
... most of the time use practical representation, if doing
something that calls for it use the precise representation
lloyd: most of the time precision won't matter - only in rare cases it will
david: it does need to be
retained
... our goal is to make it computable
tony: don't want annotation to be considered for computation
david: precisiondecimal is not a subtype of decimal
eric: Pellet has bugs around this - other reasoners may have, as well - doesn't show up in SPARQL engines
david: will want to do
comparisons freuently in SPARQL
... would like to remove option to use xsd:decimal alone in
RDF, as toolkits will lose information and cannot round-trip -
could use xsd:decimal and an additional attribute
tonY: FHIR decimal is a complex type - has xsd:decimal, and could add an additional attribute - like an extension
david: best 2 options: (1) use xsd:precisiondecimal (2) use xsd:decimal, plus additional attribute for precision
erid: problem is across all of the computational platforms
claude: how will the extension helo?
david: need to capture two pieces
of information: (1) quantity (2) precision
... Tony's suggestion - use fhir:decimal, with base xsd:decimal
plus the additional attribute to convey precision
... precision is number of digits to the right of decimal
... from SPARQL query perspective, Tony's suggestion with
precision attribute probably will work best as tools will be
able to recognize xsd:decimal
... will use xsd:decimal for schema validation, but still need
to keep track of number of precision digits
lloyd: if people read the
documentation and do what they are supposed to do, they will
retain the trailing zeros, but practically most will not as
they don't read the documentation
... precision should be captured as a standard extension
... not sure how (or if) reference implementations handle
this
david: maybe we are premature discussing this for FHIR RDF?
lloyd: should initiate and pursue
this getting settled - ask the questions to implementers
... making precision a separate element makes it harder to
lose
claude: move to 1.1?
lloyd: no interest of FHIr in moving to 1.1
david: from an RDF perspective,
an additional attribute is ideal
... should we as a group provide feedback to FHIR?
lloyd: better to do that with evidence
<dbooth> https://hl7-fhir.github.io/datatypes.html#quantity
lloyd: decimal usually used in
places like sort order, where precison is not relevant - where
precision is relevant we use quantity - an additional element
can be added there with less cost
... the simple types can only add an extension
david: precision logically would fit into quantity
paul: for currency precision is implied
david: not convinced that is necessarily always true - fractional elements are sometimes used
claude: for measurement will use quantity, can't think of other cases
lloyd: analysis of use of decimal
in FHIR
... most of the cases precision will not matter
... precision is needed for quantity, sampled data, timing,
etc.
david: how do we go forward on this:
lloyd: go forward by seeing how
reference implementations handle it
... if majority of implementers get it wrong would imply that
the design we have is vulnerable and we should rethink it - the
first question is FHIR's implementation (not RDF)
tony: the FHIR spec can't retain the precision
lloyd: FHIR implmentation is xs:decimal+
<dbooth> simple test might be the try round tripping 12.000
lloyd: if implementers are doing it right in XML and JSON, may use that in RDF - if not, need to re-think it in FHIR
david: action item - check to see
if precision is preserved with round-tripping in reference
implementations currently
... xsd documentation says you can lose the precision, but FHIR
documentation says that you cannot
... we know that the FHIR documentation of decimal is not
lossy
eric: it's not even clear how queries should behave for these comparisons
david: you are asking whether the comparison rules are clear in FHIR?
paul: quick test in .net - precision digits are not retained
david; need to check the reference implemtations behavior
david: provide input - mechanism
is to gather evidence - check the reference implementations -
who will do that?
... if reference implementations are not in compliance, that is
evidence that people are likely to screw it up
tony: but may not be possible to do this with xsd:decimal
paul: if something is inherenty
wrong, we should fix it
... bet people are not handling it because they didn't read the
spec
eric: something in the spec that no one anticipates effectively isn't in the spec
<dbooth> ACTION: Lloyd to ask James and Ewot about the underlying precision retention of xsd:decimal values [recorded in http://www.w3.org/2015/07/28-hcls-minutes.html#action01]
<trackbot> Error finding 'Lloyd'. You can review and register nicknames at <http://www.w3.org/2014/HCLS/track/users>.
loyd: will take action item to ask Ewout and James how they handle precision
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) Found Scribe: rhausam Inferring ScribeNick: rhausam Present: David_Booth EricP Claude_Nanjo Lloyd_McKenzie Paul_Knapp Rob_Hausam Tony_Mallia Marc_T Found Date: 28 Jul 2015 Guessing minutes URL: http://www.w3.org/2015/07/28-hcls-minutes.html People with action items: lloyd WARNING: Input appears to use implicit continuation lines. You may need the "-implicitContinuations" option.[End of scribe.perl diagnostic output]