Stream: conformance
Topic: Conformance Resource FMM levels
Grahame Grieve (May 08 2016 at 19:45):
Here's my list of proposed FMM levels for the conformance resources for STU3:
CodeSystem 3
ValueSet 4
ConceptMap 3
ExpansionProfile 2
NamingSystem 4
StructureDefinition 4
StructureMap 1
DataElement 2
Conformance 4
OperationDefinition 3
SearchParameter 3
CompartmentDefinition 2
ImplementationGuide 2
TestScript 2
Grahame Grieve (May 08 2016 at 19:45):
comments welcome
Grahame Grieve (May 08 2016 at 19:46):
advance note: I will be proposing to drop Data Element post STU3 on the basis that logical models handle this better
Richard Kavanagh (May 09 2016 at 14:06):
I don't see how Logical Model replaces Data Element for all use cases. We are looking at Data Element as a good match for use cases around various "master file" repositories. Without DataElement I'm not seeing how we would achieve this now.
Grahame Grieve (May 09 2016 at 14:15):
how do you think they differ?
Josh Mandel (May 09 2016 at 14:18):
For TestScript: I don't see how it will meet the criteria for FMM2 ("the artifact has been tested and successfully exchanged between at least three independently developed systems"). That would mean we need 3 independently developed implementations of TestScript, right? How many do we have?
Jason Walonoski (May 09 2016 at 14:25):
There are two TestScript execution engine implementations that I am aware of: Touchstone and Crucible.
Brian Postlethwaite (May 09 2016 at 14:31):
The Dropping of DataElement would need to consider the Questionnaire SDC work where they are pointing to these. @Lloyd McKenzie might have some thoughts.
Grahame Grieve (May 09 2016 at 14:32):
I generate test scripts by request. I believe that makes me an implementation, though you might argue that is not a proper implementation
Grahame Grieve (May 09 2016 at 14:33):
and many of us write them for connectathons. I'm not sure whether that counts
Grahame Grieve (May 09 2016 at 14:33):
regarding data element - that 's a post STU 3 proposal
Lloyd McKenzie (May 09 2016 at 23:35):
Need to add Task to that list (FMM target=2). I think saying that DataElement is a logical model is going to be a stretch for a lot of folks.
Lloyd McKenzie (May 09 2016 at 23:35):
Also, we don't own DataElement - OO does
Grahame Grieve (May 09 2016 at 23:39):
why is it a stretch?
Grahame Grieve (May 09 2016 at 23:40):
and Task is not a conformance resource. it would be on a different level
Chris Grenz (May 10 2016 at 00:41):
Question - does FMM only cover the structure of the resource, or is the behavior included. For example, I'm not ok with SD going to 4 until snapshot is well specified. But, I don't expect that work to change the elements, only the behavior.
Chris Grenz (May 10 2016 at 00:43):
Separately, my 2 cents on Conformance is that it's not well suited to what (I think) is the primary use case which is to determine whether an implementation supports the necessary features to support a particular application. I just might not be hearing the broader community, but anecdotal conversations have revealed dissatisfaction with the >1 MB conformance resources that some servers are producing.
Grahame Grieve (May 10 2016 at 00:49):
none of us are pleased about the size of it. but no immediate solution to that has popped out in our dsicussions
Grahame Grieve (May 10 2016 at 00:49):
I can return a summary version on request, which is much smaller
Grahame Grieve (May 10 2016 at 00:53):
I don't think lack of exact agreement on h\the snapshot generation aligment is a reason not to got to FMM 4 unless we think there's a structural problem. But I don't think there is
Lloyd McKenzie (May 10 2016 at 01:11):
Task may not be a conformance resource, but it's a resource owned by FHIR-I
Lloyd McKenzie (May 10 2016 at 01:13):
DataElement is something that most implementers think about as representing a single piece of data. It's an entry in a masterfile for a type of observation that can be performed. Saying to use the same structure we use to define data models with is going to be hard for people to wrap their heads around right now.
Grahame Grieve (May 10 2016 at 11:22):
well, the next task is to make a list of non-conformance resources owned by FHIR-I and give them FMM targets. It certainly stimulated interesting discussion. I'll make a list
Grahame Grieve (May 10 2016 at 11:30):
Other FHIR-I resources + suggested target levels:
List 1
Media 2
Binary 3
Bundle 5
Basic 1
MessageHeader 2
OperationOutcome 5
Parameters 3
Subscription 2
Grahame Grieve (May 10 2016 at 11:30):
I'll put this on the main implementers channel
Ewout Kramer (May 10 2016 at 12:05):
Has ImplementationGuide been used outside of the build tool? "exchanged between systems" maybe not be completely applicable here, we could exchange it, but I think the IG resource should be able to be used to contain IGs that differ from the ones in the core spec. At Furore we have tried to apply it to how we generate the Norwegian IGs, but we could not make it fit.
Grahame Grieve (May 10 2016 at 12:41):
well, we should certainly talk about the details of that
Grahame Grieve (May 10 2016 at 13:01):
@Chris Grenz would you like to extend the discussion of the conformance resource? what to do to reduce it's size.... if we break it up, then won't implementers have to fetch all the bits? that would be worse...
Grahame Grieve (May 10 2016 at 13:01):
I don't think that there's things in there that people don't use....
Richard Kavanagh (May 10 2016 at 14:03):
We use the IG resource in our tooling and it almost works for us, though we use gratuitous amount of MarkDown where we probably should not.
Chris Grenz (May 10 2016 at 16:13):
@Grahame Grieve I was mulling a more generic "feature" concept for conformance that would allow an operation to check a set of features. This wouldn't reduce the size of the conformance resource, but would allow querying a server "lightly" for a small set of needed features. It would also provide a payload for OperationOutcome where a feature isn't supported.
Chris Grenz (May 10 2016 at 16:14):
Finally, it would allow definition of custom features (URI identified) without need for conformance extensions.
Chris Grenz (May 10 2016 at 16:15):
I'm not sure I'm the right person to be commenting though...seems better suited to SMART or other app devs, esp mobile where heavy conformance is especially problematic.
Chris Grenz (May 10 2016 at 16:16):
iOS/Andriod have a mature model for this...might want to look there for inspiration?
Michel Rutten (May 10 2016 at 16:17):
@Chris Grenz I can imagine that a client, instead of fetching the (possibly very large) server Conformance statement and verifying if required capabilities are supported, it could send a sparse Conformance statement to the server that expresses only the required capabilities where the server could simply respond a yes/no. Does that make sense?
Grahame Grieve (May 10 2016 at 16:25):
These are interesting ideas - I'm glad I asked the question ;-)
Chris Grenz (May 10 2016 at 16:56):
yes - that could work. Need to get the mobile devs in this discussion
Lloyd McKenzie (May 10 2016 at 19:42):
I would hope List could be at least 2 and ideally 3. It's pretty important for lots of different areas, including Documents
Grahame Grieve (May 10 2016 at 19:44):
I haven't seen it in the wild
Lloyd McKenzie (May 10 2016 at 19:47):
Really? No allergy or problem or any other lists?
Lloyd McKenzie (May 10 2016 at 19:47):
Argonaut isn't bothering with it?
Grahame Grieve (May 10 2016 at 19:47):
no. not at the moment
Ewout Kramer (May 10 2016 at 20:55):
INM has spoken about moving the "Event" part of Conformance out to a MessageEventDefinition conformance resource (analogous to OperationDefinition), so this means we might not want Conformance at FMM4?
Lloyd McKenzie (May 10 2016 at 20:56):
FMM 4 requires full validation against scope and I don't think Conformance has been well-exercised against anything other than REST. Messaging is particularly weak.
Grahame Grieve (May 11 2016 at 10:23):
so we have a number of issues around conformance. Do we try to address them? or let it slide?
Ewout Kramer (May 11 2016 at 15:59):
No, let's turn this into a topic
Last updated: Apr 12 2022 at 19:14 UTC
