Stream: conformance
Topic: FHIR registry question
Grahame Grieve (Dec 15 2017 at 05:09):
Hey @Ewout Kramer etc at Furore - GF#13648 - what is the right link for this?
Michel Rutten (Dec 15 2017 at 09:58):
Do you mean a link to all the CodeSystems in the FHIR registry?
https://registry.fhir.org/search?q=resourcetype:codesystem
Grahame Grieve (Dec 15 2017 at 10:57):
that's all the naming systems
Ewout Kramer (Dec 18 2017 at 13:59):
So, https://registry.fhir.org/search?q=resourcetype:namingsystem is the answer. Though if you do that, you'll get the rather ugly server-assigned ids - that must be a consequence of the fact that namingsystems don't have a canonical url.
Ewout Kramer (Dec 18 2017 at 14:01):
Am I correct, @Vadim Peretokin ?
Vadim Peretokin (Dec 18 2017 at 14:05):
That's correct!
Vadim Peretokin (Dec 18 2017 at 14:05):
No canonical URL, no narrative, so it doesn't look great.
Michel Rutten (Dec 18 2017 at 14:40):
IMHO all conformance resources should have a canonical url.
Ewout Kramer (Dec 18 2017 at 14:43):
I've lost that discussions too many times. Many seem to agree, but when I argue it at a FHIR-I call, I am all alone.
Lloyd McKenzie (Dec 18 2017 at 15:33):
The point of a canonical URL is to support referencing from other resources. NamingSystems aren't referenced - except by Coding.system and Identifier.system. And the URLs sent there are not canonical URLs in the same sense. It's possible for multiple URLs to be active simultaneously and NamingSystem lists all of them. It marks one as "preferred", which is as close as you get to a canonical URL
Grahame Grieve (Dec 18 2017 at 18:17):
you can mark something as preferred that is not a URL.
Grahame Grieve (Dec 18 2017 at 18:17):
just give it a canonical URL
Lloyd McKenzie (Dec 18 2017 at 18:42):
"preferred" is a flag that applies once each for URLs, OIDs and v2-style naming systems. Giving it a canonical URL has no purpose and will invariably cause confusion. It is always findable by a URL. That doesn't mean the notion of "canonical" URL is meaningful for it.
Grahame Grieve (Dec 18 2017 at 18:54):
it's causing confusion by everyone who's actually implemented it that I know
Lloyd McKenzie (Dec 18 2017 at 19:19):
Adding it is going to create even more confusion - because then you're going to have situations where the canonical URL isn't the preferred URL - because the preferred URL is mutable and the canonical URL isn't
Grahame Grieve (Dec 18 2017 at 19:31):
that's exactly the benefit
Lloyd McKenzie (Dec 18 2017 at 19:35):
But why force a URL that's not actually a legitimate URL for the system?
Lloyd McKenzie (Dec 18 2017 at 19:35):
The preferred URL might be "urn:oid:1.2.3.4" and there are no other URLs
Lloyd McKenzie (Dec 18 2017 at 19:36):
Introducing a canonical URL means you suddenly have to have a proper URL, but it's not going to be one of the listed system ids - which is going to confuse the heck out of everyone.
Grahame Grieve (Dec 18 2017 at 19:37):
so the point is: everyone is already confused because it's not there. And that makes for all sorts of special cases around naming system since it's the only that doesn't have it. So if people are going to be confused, let's make it consistent
Grahame Grieve (Dec 18 2017 at 19:37):
as things stand now, there's no identification of the registration. It's not that hard a concept to understand, actually
Lloyd McKenzie (Dec 18 2017 at 19:45):
Then we need to make it Extremely clear that you never look at the canonical URL for what to put in Coding.system or Identifier.system. You always look in NamingSystem.uniqueId.value. (On occasion the canonical URL might show up as one of the uniqueId.values, but this isn't expected or even common practice.)
Simone Heckmann (Dec 20 2017 at 15:23):
+1 for adding canonical url
I am one of the confused people.
Last updated: Apr 12 2022 at 19:14 UTC
