Topics discussed:

• From the connectathon: a use case for FHIR servers using ViewDefinitions to denormalise the views applications actually need — dashboards and reads that are painful to assemble out of FHIR search. That wants close to unlimited FHIRPath, resolve() included. The counter was to keep the subset as the base and layer resolve() and friends into an optional advanced mode rather than the required core. Expected to come back.
• The IG publisher now has a ViewDefinition runner in its code base, used for analytics over IGs themselves — how many code systems, how many codes — and on the International Patient Summary to compare profiles from different countries. When an IG is published it ships a SQLite database of its content that anyone can query with a SQLite client.
• ViewDefinition simply becoming a resource in R6 was floated. The objection was timing: nobody knows when R6 lands, so jumping the queue to it skips the stage that would actually help — an incremental status that gets more eyes and more use cases on the spec sooner.
• Ballot for comment had been suggested as a way to put the spec in front of people who do not attend the weekly call without entering formal standards development. FMG pushed back: too much is going through ballot-for-comment already, and this one is expected to become a real standard anyway. The advice was taken, though the goal of reaching a wider audience stands.
• MIMIC-IV on FHIR is being used as a test vehicle — around 500 million resources of semi-public de-identified real inpatient data, as opposed to everyone being sick of Synthea. Driving the full set through a view runner surfaced validation problems and a transformation process that does not scale, so it is being reimplemented with validation built in. Separately, the functional model write-up is close to publication: implementers report it gives a different lens on the spec — read the four functions, see how others approached them, and you can port an implementation to your own language.