Stream: cds hooks
Topic: Reliable hook permalinks
Lloyd McKenzie (Apr 14 2020 at 17:25):
The CRD specification has references to a number of hooks, but none of the ones we need are actually published as a 'stable' balloted specification. Our challenge is that when we formally publish the CRD spec, we need to have a permalink that refers to the hook definition (and ideally that points to a 'frozen' view of the hook definitions). With the HL7 core spec, we create frozen snapshots for connectathons and ballots every 4-5 months and - occasionally - we create special ones for a particular publication. However, there doesn't seem to be an equivalent process for the CRD spec. We really need something like this. We can't reasonably hold off CRD until every hook we care about passes STU and we can't publish a spec where the links either break or the names of parameters, cardinalities or even types of what's linked to differ from what the specification assumes.
So: can we institute a process where the draft CDS Hooks spec (including hook definitions) can be occasionally snapshotted and those snapshots made available at a permalink site? E.g. http://hl7.org/fhir/cds-hooks/2020-May or something like that?
@Isaac Vetter @Josh Mandel @Grahame Grieve
Bryn Rhodes (Apr 15 2020 at 00:04):
We just balloted the patient view hook, doing exactly this: https://cds-hooks.hl7.org/hooks/patient-view/2020Feb/patient-view/
Isaac Vetter (Apr 16 2020 at 15:30):
Lloyd, you're saying regardless of any given maturity, we just want a frozen snapshot of a hook, right? Github has excellent versioning, would something like this work?
Lloyd McKenzie (Apr 16 2020 at 15:52):
Do those versions ever go away/get collapsed? It may be viable, though a frozen snapshot with a distinct URL (e.g. as per http://hl7.org/fhir/2020Feb) would be better. @Grahame Grieve, thoughts?
Isaac Vetter (Apr 16 2020 at 18:55):
the above url will never go away, never get collapsed (it's github exposing the underlying git version control). (It's a bit awkward pointing to a non-HL7 url, and the visual UI isn't perfect, but those are the only problems).
Grahame Grieve (Apr 16 2020 at 20:16):
but those are real problems
Lloyd McKenzie (Apr 16 2020 at 20:33):
Is there a problem taking frozen snapshots as we do for the core spec and FHIR IGs?
Bryn Rhodes (Apr 16 2020 at 20:39):
The only problem I can see is that our only mechanism for doing that is as an HL7 project as part of a ballot.
Lloyd McKenzie (Apr 16 2020 at 22:08):
It's an HQ function. Grahame and Lynn can take snapshots of a site at any time - though it's not a 2-second process, so not something they'd do without a formal request and a good reason.
Lloyd McKenzie (Apr 16 2020 at 22:09):
Examples of such have historically been coming connectathons
Bryn Rhodes (Apr 16 2020 at 22:16):
Okay, so are you wanting to support the CRD hooks for the upcoming connectathon?
Bryn Rhodes (Apr 16 2020 at 22:18):
Or no, reading your initial post, you're wanting to publish. Which hook definitions are required specifically?
Lloyd McKenzie (Apr 21 2020 at 18:21):
Sorry @Bryn Rhodes, missed this question. The hooks used by CRD are appointment-book, encounter-start, encounter-discharge, order-select, order-sign.
Lloyd McKenzie (Apr 21 2020 at 18:21):
But it's not just the hook definitions. Presumably the specification itself should be snapshotable
Bryn Rhodes (Apr 21 2020 at 21:16):
And it is, but we consider the specification and the hooks separate, so they can publish/ballot independently.
Lloyd McKenzie (Apr 21 2020 at 21:19):
Sure, but a snapshot should ideally be of both
Lloyd McKenzie (Apr 30 2020 at 17:11):
Bumping this thread - is creating frozen snapshots of the CDS Hooks spec (and the hooks) similar to what's done with the core spec something we can do? I'd really like to get the hyperlinks in CRD nailed down... @Bryn Rhodes @Isaac Vetter
Bryn Rhodes (May 01 2020 at 17:18):
So, I have a snapshot ready to go for this, but I noticed a new PR against appointment-book today?
Bryn Rhodes (May 01 2020 at 17:20):
@Lloyd McKenzie , do those changes to appointment-book need to be in this snapshot?
Lloyd McKenzie (May 01 2020 at 17:53):
Don't know that they "need" to be. It'd be sort of nice if they were - part of CRD ballot reconciliation and requested by Isaac, but it doesn't impact the function of the CRD spec.
Lloyd McKenzie (May 27 2020 at 16:25):
Following up on this - @Bryn Rhodes @Isaac Vetter
Bryn Rhodes (May 29 2020 at 15:12):
@Lloyd McKenzie , I am working on a snapshot that will include all the currently non-deprecated hook definitions. Hoping to have that ready today, and I will pull from the latest version of all the hooks on the github.
Matt Varghese (Sep 09 2020 at 17:00):
Isn't there a bigger question here of some standard proposal using a hook that does not have an appropriate maturity level? If the referencing proposal is standardized, that becomes a de-facto standardization of a hook which, per CDS Hooks community, is not sufficiently mature?
Lloyd McKenzie (Sep 09 2020 at 18:21):
There'd be a limit to how far the standardization could reasonably go. You couldn't have a normative part of a spec pointing to a non-normative hook. However, the process of getting hooks into ballot at all is 'slow', so not being able to reference a reliable location from an STU specification is problematic.
Last updated: Apr 12 2022 at 19:14 UTC
