Stream: implementers
Topic: OID vs. URL for Identifier and Code Systems
Lloyd McKenzie (Mar 31 2017 at 19:03):
Canada is working on a policy document for identifier and code systems - what the URIs should be, what should be registered, where it should be registered, etc. Most of the document is not Canadian-specific and may be of interest to others. Feel free to add comments/provide feedback. And don't be afraid to challenge assertions you think are out-to-lunch. If there's general interest, we could also look at migrating this to a wiki page and potentially pointing to it from a future version of the FHIR spec. https://docs.google.com/document/d/1ClF3XpIrE2FiIDG7K7VD3pPjOPx3T950Sfi1Ip8ORcU/edit#heading=h.641auhfi9fvs
Eric Haas (Apr 01 2017 at 18:55):
This would actually make this topic a lot easier to navigate. Thinking of my experience with the CDCREC codes was very ad hoc placing calls to CDC to try to find the right person etc and eventually using a URN since we ran out of time. ... But isn't there an HL7 project to address some of this stuff already? I don't know where that has landed.
Lloyd McKenzie (Apr 01 2017 at 20:40):
I'm not aware of a project to deal with this. If there is one, pretty sure it didn't come through MnM - which I would have expected.
David Hay (Apr 02 2017 at 00:12):
I think this is a great idea, and will follow (and possibly comment) with interest...
Robert McClure (Apr 02 2017 at 15:57):
@Lloyd McKenzie The document does not indicate that it is focused on Canada. What part is "Canada-specific"?
Lloyd McKenzie (Apr 02 2017 at 16:33):
Look for the section "What should the root be/where to register"
Alexander Henket (Apr 02 2017 at 21:00):
The short version of that for us is that we extended our OID Registry with a new column. Assumption is that if that is populated, it is then the preferred URI -- for FHIR. If a new system comes up (didn't happen so far), it shall also have an OID. The only reason for URIs instead of OIDs is readability (readability also is constant reason for debate... very deep sigh), so we try to be as readable as possible. If that makes for a lengthy URI, then so be it; we imposed no max so far.
Alexander Henket (Apr 02 2017 at 21:24):
Interestingly enough system URIs turn out to be as resolvable as OIDs. If you don't register them properly in one or more recognized registries: you will not resolve them. A tiny error in the URI string can be just as killing as one in an OID.
Lloyd McKenzie (Apr 02 2017 at 21:29):
We'd want our URIs to be valid FHIR URLs, which imposes a limit of 64 characters on the "id" portion of the OID. (Speaking of which, I should note that in the paper :))
Lloyd McKenzie (Apr 02 2017 at 22:12):
The difference is that if you make a typo in an OID, you can end up with a "valid" OID, so even if you have checking to make sure that OIDs are "legitimate", the problem won't get caught. That issue is much less likely (and much eaiser to detect) with URIs. OIDs are never resolvable - looking something up in a registry isn't resolution. URLs are resolvable by using the DNS system and hitting a server. If you have to fall back to a registry, then there's no significant difference other than that FHIR registries are ususally RESTfully searchable. OID registries, not so much - at least not historically.
Alexander Henket (Apr 03 2017 at 03:47):
OIDs were poorly registered in jurisdictions other than ours, true. Typos in our jurisdiction are always caught in the qualification process. We check every level and audit their implementation strategy.
.
The net effect of a typo is that you cannot process that codesystem. Even in URIs it can be subtle. Would you know there's a typo in this one? https://referentiemodel.hng.org/tabellen/nhg-tabel-45-diagnostische-bepalingen. Identifiers is even more tricky, because it may go undetected for a longer period if the vendor makes a typo in setting the base at a new customer.
.
Anyway. The main point of the document is to stress what they are, how important it is you get them right, and what "get them right" means. I think it's a great initiative. I already added a number of comments that are hopefully helpful, and linked to it from our best practices wiki.
Eric Haas (Apr 03 2017 at 16:40):
@Lloyd McKenzie I was thinking of the "FHIR Repository Requirements Project". But that seems broader than this document.
Lloyd McKenzie (Apr 03 2017 at 17:05):
Yeah. That's about what the registry should do. This paper is more about what should be put in the registry.
Last updated: Apr 12 2022 at 19:14 UTC
