ShEx validator service

eric: harold and michael are using it. they're also using Jose's implementation. harold and grahame are putting jose's implementation into the build spec.

dbooth: will that do round-trip testing?

eric: no.
... We could try writing some genx in there and do round-tripping.
... But not needed for round=-tripping, becauise grahame does it through their reference library anyway.
... Harold has validated all 450 examples in the FHIR core!
... And it found a bunch of errors.

dbooth: any need followup?

Validation use cases

eric: Three variables: 1. whether or not you are dealing with just the core, vs extensible stuff (FHIR extensions); 2. looking at a single resource on its own, vs validating a constellation of resources. 3. Are you dealing with profiles or only the core FHIR spec?
... And another consideration: where you have a profile, but there are extension in the data that you do not know about from the profile, in particular non-modifier extensions.
... E.g., observation for a Mayo clinic admission, and it is also a CDC-monitored respiratory disease observation
... We need different constraints depending on these criteria.
... If you want to enforce the validity.
... If something references a fhir:MedicationDispense, do we validate that referent?
... As soon as you say that you want to chase all of the referents, then you pull in all of ShEx at once.
... But if you don't pull in the referent then the validation will fail.

dbooth: Don't want to have to chase everything. Maybe you should have to say explicitly what you want to validate.

eric: Could have a skinny mode (that only validates the current resource) and a fat mode (that validates everything that is referenced, recursively)

dbooth: The fat mode sounds like a high bar, because it requires everything to be loaded.

eric: yes, but that's a common situation, because usually have the full patient record there.

shahram: we have a similar problem, but a bit different. We create list of known resources, and stop there.
... We treat the known resources differently -- they are resources of a known type.
... We create a list of the types that we want to validate.
... We want to aggregate clinical data.

dbooth: how about if the reference declares the fhir:intendedType . Then the rdf:type would be optional, but if it is there, then it must match the fhir:intendedType ?

eric: MedicationOrder references an Observation, and an rdf:type triple that says it is an Observation.
... Or instead it could say fhir:intendedType
... The local validation just says that an Observation needs an rdf:type Observation.
... And the extended validation also needs a status. But you cannot tell what the status is if you just see a reference to an Observation.
... Are people using XML schema for validation?

amol: We use xml schema.

eric: Are you mostly validating workflow? Amol: yes.
... According to grahame, the xml does not validate any of the references.
... Though in theory it could look for the string "Patient" in the URL
... Maybe we should do the skinny validation first, using a different type for the reference than the referent.
... What RDF advantages are you trying to get?

amol: We can handle many kinds of messages. We have clients who will send CSV files.

eric: Other axes: 1. constellation vs local 2. whether or not it has FHIR extensions; The model Josh and I wrote a couple of years ago modeled extensions as more RDF statements. The downside is that we cannot close the shapes, i.e., disallow anything else. 3. Is it dictated by a profile vs base FHIR? Is there a profile that say what else to expect in the Observation?
... Also, how much will people validate with the expectation of a closed shape? i.e., what to do about extra data? Is that a viable way to encode extensions?

Failing (negative) validation tests

eric: Grahame thought he could do something.
... He has a dir of tests, and it has both negative and positive tests. He thinks he will be able to produce them, and in each format.

This is scribe.perl Revision: 1.144  of Date: 2015/11/17 08:39:34  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/4 categories: (2x2 matrix)/Three variables:/
No ScribeNick specified.  Guessing ScribeNick: dbooth
Inferring Scribes: dbooth
Present: David_Booth EricP Rob_Hausam Brian_Pech Shahram Amol James_Anderson
Found Date: 07 Jun 2016
Guessing minutes URL: http://www.w3.org/2016/06/07-hcls-minutes.html
People with action items: