The way to spread SQL on FHIR is a slow, naive implementation inside HAPI — Apr 21, 2026
John — distribution, not outreach
The answer to how do we get more people involved wasn't outreach — it was distribution.
There's already an implementation in FHIR core, which means it's reachable from HAPI, and the same could be done for .NET.
It would never set the world on fire on performance — a naive implementation over HAPI's JPA schema is a baseline, not a contender. But that's the point: it's a gateway drug. Make the functionality work inside what people already run, and the ones who need speed will go looking for an analytics-focused implementation on their own.
The same argument extended to bulk data
Plenty of people tried to bolt bulk export onto the side of a transactional FHIR server and it didn't work. The question is where those people go next, and whether SQL on FHIR has anything to offer them.
Gino supplied the opening: there's an Argonaut project this year reconciling bulk data, bulk import and bulk publish into one IG, ending in a publication cycle. If the group wants notes about SQL in that spec, that cycle is the window, and it's sooner rather than later.
How John runs connectathon tracks
More organised than most: a participant capability matrix filled in ahead of time, one row per system, so on the day you can look at the table and know who to talk to for a client.
Applied here, the split is between people who want to pick up a runner and run ViewDefinitions on their own platform — self-contained — and people who want to poke at somebody else's server, which only works if the endpoints and auth details have been harvested in advance.
Craig — the best analogy for the current moment
Semi-technical people are now building hyper-personalised apps and dashboards for themselves, and it reminds me of the era when a cheesy Visual Basic app somebody knocked together quietly became load-bearing for half a company's business process.
Steve had a live example from the other direction: an enthusiastic interventional cardiologist turning up with an implementation of profile validation, which is not a small thing to write, and asking whether he could contribute it.