Stream: implementers
Topic: Thoughts on External References
Joe Jordan (Oct 22 2021 at 16:50):
Curious as to what most implementers are deciding on allowing external references on their servers. My thoughts initially were that we should simply use relative references when creating resources, i.e. "Patient/{id}" rather than "https://domain.com/fhir/Patient/{id}", however I was interacting with an external server recently and the fhirUser assigned to the login is a full URL reference, and it got me thinking. Perhaps for data portability purposes, ALL references should be saved as a full URL, so as to not provide any ambiguity to any external servers that may ingest our data as to where those references can be found/originate from.
Our HAPI FHIR server is spun off of the hapi-fhir-jpaserver-starter project, and I believe the default is to not allow external resources. Is this common practice across production implementations? Are there any concerns surrounding data portability in leaving off the originating source of references? Am I overthinking this topic too much? :grinning: Appreciate any advice, TIA.
Lloyd McKenzie (Oct 22 2021 at 16:56):
If you ingest something that has a local reference and you don't also ingest the thing referenced, then you need to turn the local reference into a remote reference. Sticking with local references makes it easy if you're cloning sets of data from one server to another because then you don't have to change references. It also means you don't break digital signatures (though that's not a common thing).
Last updated: Apr 12 2022 at 19:14 UTC
