Slides from Harold for today's discussion: https://lists.w3.org/Archives/Public/www-archive/2017Nov/att-0014/FHIR_conneg_v2.pdf
rob: notes from Dev Days?
harold: Yes, a recording is available.
dbooth: Notes at http://www.w3.org/2017/11/15-hcls-minutes.html
harold: Recording is at
https://www.aanmelder.nl/i/mail-click-event?id=X0XEwrSCP_IwvnsgIVyAxShJMzIyODQ2MwpTJ1x4ZmRceGJjXHg5N1x4ZGFceDE2XHhkMDlceGNlXHhjZEBceGM4bHBceGE3XHhlNi9ceDk0XHgwNDtceGY2XHgxNlx4YjhtXHg5YVhceGI0Ylx4Yzd7XHhmZVx4ZWZceGEwJwpwMQp0cDIKLg.
... but it requires a password, because it is for Dev Days
participants. I have the password.
harold: Tutorials were well
received, and people have started trying some interesting stuff
already.
... Someone from the NL used python scripts to convert a
thesaurus.
... It was Johan van Soesrt (Maastrick)
vocabularies for codesystems
rob: Discussed it a bit at Dev Days, but still not done.
harold: I got permission from Grahame to drop the OWL stuff into the build.
rob: Just needs to be implemented. I could help with it.
harold: Grahame wants to completely revamp the W5 ont.
dbooth: What motivation?
harold: He wants to add more
semantic content to FHIR -- organize and categorize FHIR
resources.
... Also, when we inherited fhir.ttl it had stuff that didn't
work in OWL. We decided to include stuff to make it work in
OWL. But Grahame really wants the code systems and valuesets
back into it.
ISSUE: Need to get code systems and valuesets back into fhir.ttl
<trackbot> Created ISSUE-47 - Need to get code systems and valuesets back into fhir.ttl. Please complete additional details at <http://www.w3.org/2014/HCLS/track/issues/47/edit>.
(Harold reviews slides at https://lists.w3.org/Archives/Public/www-archive/2017Nov/att-0014/FHIR_conneg_v2.pdf)
harold: I wanted to support more
than one RDF format
... recommendation: Recommendation 1
We don’t need to separate OWL Resource instances from
non-OWL instances
AGREED: There will be a single file w/ OWL ontology declaration.
harold: Recommendation 2
_format in the URL overrides the accept header.
• Documentation on http://build.fhir.org/http.html will be
updated and clarified
• “must” / “may” and “must not” will be made clear in the
spec wrt. _format=ttl, format=text/ttl, ...
• rules willl be added to a conformance testing tool.
(all in agreement)
harold: Recommendation 3
_format redirect rules will be made explicit
• _format must be included in URL rewrite
• rules willl be added to a conformance testing tool.
(all in agreement)
harold: On MIME types, we need to
enumerate them and say which ones MAY/MUST be handled.
... Recommend
http://<server>/CodeSystem/account-status/owl — returns
OWL
representation
eric: Accept header allows you to ask for variations on a resource. The challenge is that when you have the same info both in FHIR valueset vs OWL valueset, how to negotiate to differentiate them?
harold: But in this case it is different content.
dbooth: But one has metadata, like domain and range.
eric: There was a workshop at W3C
about how to specify media type profiles -- Accept-Profile
header was proposed. This would allow e.g., specific RDF
vocabularies to be requested.
... In Apache I would have all headers and could switch on the
profile in addition to other things.
harold: Propose: http://<server>/CodeSystem/account-status/owl — returns OWL
representation
AGREED: http://<server>/CodeSystem/account-status/owl/version/1.0 — returns OWL representation
harold: Problem: Protege uses Accept header: application/rdf+xml, application/xml; q=0.5, text/
xml; q=0.3, */*; q=0.2
scribe: This causes FHIR/XML to
be returned instead of Turtle.
... Proposal: if the Accept header is that exact header, then
return Turtle instead, assuming that it is protege asking for
it.
dbooth: What about checking the User-Agent header, since this is a protege glitch?
eric: How often will we want to demo this before we get protege fixed?
harold: If protege resolves itself before we reach recommendation, then we can change it.
eric: Do we want to conneg between the FHIR and OWL versions?
dbooth: I don't think we need to conneg between them. So they can just use different URLs.
harold: Also, a PUT of a FHIR resource in OWL will not go back to FHIR.
eric: When you GET the structure def of a valueset, and you also want to GET the corresponding OWL, how do you know what is the URL for it?
harold: I don't think we need to use conneg between FHIR resource and OWL ont.
dbooth: I agree.
<scribe> ACTION: EricP to followup with Protege community and developers to fix their Accept header for OWL formats
<trackbot> Created ACTION-93 - Followup with protege community and developers to fix their accept header for owl formats [on Eric Prud'hommeaux - due 2017-12-05].
<scribe> ACTION: DBooth to find out user agents for Protege, TQ, Stardog, Jena
<trackbot> Error finding 'DBooth'. You can review and register nicknames at <http://www.w3.org/2014/HCLS/track/users>.
AGREED: special case protege using User-Agent header (preferably) or Accept header (if needed)
eric: Clarifying: protege and other tools that do explicity accept Turtle.
<inserted> dbooth: eric, Protege sets the Accept header here: https://github.com/protegeproject/protege-owlapi/blob/47da17bd5a962f276d455cc2272beb771aa86add/src/main/java/org/protege/owlapi/util/IOUtils.java#L61
ADJOURNED
This is scribe.perl Revision: 1.152 of Date: 2017/02/06 11:04:15 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: Irssi_ISO8601_Log_Text_Format (score 1.00) Succeeded: i/ADJOUR/dbooth: eric, Protege sets the Accept header here: https://github.com/protegeproject/protege-owlapi/blob/47da17bd5a962f276d455cc2272beb771aa86add/src/main/java/org/protege/owlapi/util/IOUtils.java#L61 Present: David_Booth Harold_Solbrig Rob_Hausam EricP No ScribeNick specified. Guessing ScribeNick: dbooth Inferring Scribes: dbooth Found Date: 28 Nov 2017 People with action items: dbooth ericp WARNING: IRC log location not specified! (You can ignore this warning if you do not want the generated minutes to contain a link to the original IRC log.)[End of scribe.perl diagnostic output]