guidance on client caching questionnaire content

[kjohnsen-epic]

Related to some of the language on https://build.fhir.org/ig/HL7/davinci-dtr/OperationDefinition-questionnaire-package.html-
We were designing a provider side data model for DTR, and the example $questionnaire-package responses I’ve been looking at so far have been from 5 to 500kb, which is more than enough that I wouldn’t want to simply store everything.

To deal with that I’ve been trying to understand what message content is safely cacheable / reusable. The ones I’ve been considering so far are:

• If I see Questionnaire URL + version + status=active, I think it’s safe to store that only once and reuse it in future exchanges, since services can’t change the content at that point. For example I store “MCG Knee Replacement v2” once instead of every time.
• Same for Library URL + version + status. For example I store “FHIRHelpers v4” once instead of every time.

Are my assumptions correct there, and should we include guidance to this effect?

Are there other cacheable pieces?

[kjohnsen-epic]

Talking to Bryn, he mentioned we should have guidance, but it needs to be taken on a case-by-case basis:

1. Questionnaires may be modular, meaning they potentially have different assemblies depending on whether the references are versioned
2. ValueSets may be definitions or expansions (or both, and potentially multiple expansions) so we need a way to both distinguish and assure fit
3. Libraries may be computable, executable, or both