Use of ECA rules over composite for KNART workflows.

[preston]

This is part question, and part commentary on composite KAS. It's not yet clear what is actionable, but at the very least I'd like to understand the rules of thumb used for authoring.

Looking over o8, an ECA KNART, it seems:

• o8 is basically just a glue structure that is intended to tie other KNARTs together, which is what a composite does via a de facto "router" element.
• Apparently due to lack of suitable ANF encoding for chaining of specific KNARTs -- e.g. "Request KNART XYZ be performed" -- the "triggers"/actions can't actually being encoded at this level. I'm not clear if this will be consistent over all ECA rules, but surmise it's a common issue.

Looking over (the unrelated) b57, a composite, it seems that:

• The ELM switch/case logic basically does the same type of thing as the embedded router artifact of o8.
• The use of trigger-related elements, while not ANF, is naturally better suited to directly chain KNARTs in the manner intended with o8.

If this is a common similarity, it feels like composites and ECAs are both fundamentally very similar, with composites being more powerful due to the ability to both make ANF assertions and invoke specific artifacts at the knowledge or wisdom level. Some questions with possible actions:

1. How was it determined when to use an ECA versus a composite, such as in cases where actions aren't encodable?
2. Will all ECAs exhibit the limitations of event and action expressiveness as o8?
3. Has superceding the existing ECA type with the more powerful composite type been considered at the KAS level? (If so, what was the resolution and recommendation?)

[WendelynB]

Technical Review.

[preston]

Thinking of the DIKW model, does it event make sense to people to be modeling these sorts of relationships at the ANF level, in which case I suppose this would be a request for invocation of a system task, or should composite workflows philosophically be treated as high-level concepts? My inclination is the latter, but very curious what others think!

[csp4z]

1. How was it determined when to use an ECA versus a composite, such as in cases where actions aren't encodable? Per discussion with Andrew & Preston, this was defined in PWS according to Cognitive.

2. Will all ECAs exhibit the limitations of event and action expressiveness as o8? Yes. a) One KNART cannot access another, b) One KNART can't invoke another - No means of functionality integrating these.

3.Has superceding the existing ECA type with the more powerful composite type been considered at the KAS level? (If so, what was the resolution and recommendation?) Yes, this is being discussed at HL7 Composite Meetings.

[csp4z]

Preston requested assignment on this.