Stream: conformance
Topic: GF#8723
Grahame Grieve (Feb 16 2017 at 21:51):
I spent an hour yesterday trying to get GF#8723 working, and ended up backing out. It would be a forward reference to another element, and we haven't had one of them before. So it's going to miss the boat.
Grahame Grieve (Feb 16 2017 at 21:51):
but while I was doing this, I was also trying to make sense of "the usecase where extensible bindings need information about bindings in the parent to correctly decide whether a concept is already present in the ancestry of bindings"
Grahame Grieve (Feb 16 2017 at 21:52):
I think that I understand the vocab validation space pretty well, but I have no idea what that means.
Grahame Grieve (Feb 16 2017 at 21:52):
which really makes it quite the extension kind of thing, in my experience.
Grahame Grieve (Feb 16 2017 at 21:54):
and made me wonder: a much more achievable technical solution to this, which seems to have the same outcome, is to add another extension like maxValueSet and minValueSet - extensibleValueSet - that goes on a binding, so you can list and track all the inherited extensible value set bindings if you really want. I'm pretty sure that you only have a need (for that whaterver-that-is- above) for extensible value sets if the binding is extensible, and the inherited bindings are extensible.
Lloyd McKenzie (Feb 17 2017 at 02:41):
Parent says Red, Green, Blue, Yellow; extensible
Child says Red, Blue; extensible
Grandchild tries to say Red, Blue, Emerald Green
Lloyd McKenzie (Feb 17 2017 at 02:41):
If all you have is the child, not the parent, you can't see that Emerald Green is invalid
Grahame Grieve (Feb 17 2017 at 02:58):
i'm confused here - which is profile, and what is the instance?
Lloyd McKenzie (Feb 17 2017 at 03:56):
This is all profiles
Lloyd McKenzie (Feb 17 2017 at 03:57):
Though we could also have the grandchild be an instance
Grahame Grieve (Feb 17 2017 at 04:03):
I feel as though the Child is wrong here
Rob Hausam (Feb 17 2017 at 04:10):
We can imagine these scenarios (using colors as examples). But do we actually need them in real implementations? And if we do, are there any simpler approaches to accomplishing the needed result? I'm perfectly willing to be convinced, but I think I would like to see (or come up with) some more realistic examples of where it is actually going to help.
Lloyd McKenzie (Feb 17 2017 at 04:13):
It's legitimate to constrain out certain concepts. Imagine something more realistic. Base value set is "all SNOMED disease codes, extensible". Child value set is "all SNOMED disease codes relating to lungs, extensible". Definitely use-cases for doing that sort of thing. And sending a custom code for "broken arm" would clearly be prohibited.
Grahame Grieve (Feb 17 2017 at 04:27):
I don't understand the child value set in this case. what is it trying to achieve?
Lloyd McKenzie (Feb 17 2017 at 05:01):
It's saying that in this profile, we should only have drug diseases - but it's possible some of them might not be in SNOMED, so it's extensible
Lloyd McKenzie (Feb 17 2017 at 05:01):
It's consciously constraining out all of the SNOMED codes that are for non-lung-related conditions
Grahame Grieve (Feb 17 2017 at 05:10):
and discarding conformance information. I feel as though this is problematic
Grahame Grieve (Feb 17 2017 at 05:11):
anyway, so far I as I can figure, this only applies when focus definition is extensible, and when parent is extensible. if focus is required, or parent is example/preferred, then this is not a ting
Grahame Grieve (Feb 17 2017 at 05:12):
so we can define inherited-extensible-valueset extension, and migrate that across the profiles, and that does what is needed
Lloyd McKenzie (Feb 17 2017 at 05:43):
That should work
Grahame Grieve (Feb 17 2017 at 05:50):
great. I
Grahame Grieve (Feb 17 2017 at 05:50):
'll do that
Last updated: Apr 12 2022 at 19:14 UTC
