FHIR Chat · Contents of a message bundle

Stream: fhir-messages

Topic: Contents of a message bundle

Vassil Peytchev (May 12 2020 at 18:47):

Currently, for bundles of type document, there is an explicit requirement that all resources in the Bundle must be reachable from or can reach the Composition resource. There doesn't seem to be such a requirement for bundles of type Message, so, theoretically, I can have a bunch of unrelated resources and still call it a message.

Also, the MessageDefinition resource allows zero or more graph definitions to exist for a given message. Does that mean that a message instance must have one and only one of the graphs as its content, that it can have any combination of the graphs as its content, or that it must have all graphs as its content?

How are these capabilities/requirements or lack thereof supposed to work together?

Lloyd McKenzie (May 12 2020 at 19:56):

We've approved a change request to add the same rule for messages

Lloyd McKenzie (May 12 2020 at 19:57):

I think we also approved changing MessageDefinition to only allow a single GraphDefinition, though I can't swear that happened. (If not, feel free to submit a new request, as that's certainly an appropriate change to make.)

Vassil Peytchev (May 12 2020 at 20:00):

I can see a possible value in having multiple graph definitions, where the only interpretation is that an instance must use one of them.

Vassil Peytchev (May 12 2020 at 20:31):

Lloyd McKenzie said:

We've approved a change request to add the same rule for messages

Ah, yes, FHIR#25257...

Vassil Peytchev (May 12 2020 at 21:06):

Lloyd McKenzie said:

I think we also approved changing MessageDefinition to only allow a single GraphDefinition, though I can't swear that happened. (If not, feel free to submit a new request, as that's certainly an appropriate change to make.)

FHIR#27123

Lloyd McKenzie (May 12 2020 at 23:43):

I can't think of a situation where a single message structure would have multiple alternative graphs that couldn't be (and shouldn't be) expressed as a single GraphDefinition.

Eric Haas (May 13 2020 at 04:20):

That is why I proposed and @René Spronk added the mustsupport to the GD redesign proposal. You can have several variants on a Bundle depending what you have.