Drasil
Drasil copied to clipboard
JacquesCarette
QDefinitions need ConceptDomains
Regarding code/drasil-lang/Language/Drasil/Chunk/Eq.hs QDefinition's cd field,
cd is intended to be manually set. Once ModelKind is implemented, as a second step, we'll want to change the combinators to require the 'concept domain'. But first, we'll have to make sure that the concept domain(s) of all of our examples exist!
The best way to deal with that is probably to create multiple new issues to deal with that. First some issues based on changing each new combinator to accept domains, then issues around adding the missing domains, then issues around fixing the examples to use these new combinators, and finally issues to remove the old combinators. So this can easily spawn a dozen new PRs.
Originally posted by @JacquesCarette in https://github.com/JacquesCarette/Drasil/pull/2303#discussion_r599603906
Which part do you want explained / feel like it needs an explanation?
Sorry, I wrote that question as unclear as it was because it was meant for me. I meant: what do we get out of attaching a concept domain to QDefinitions? With ConceptInstances, we're able to make 'concepts' Assumptions vs Goals vs other things in the SRS. What does it do for QDefinitions?
I'm not quite sure that I fully understand the intent behind ConceptChunk, ConceptInstance, and ConceptDomain.
The concept domain is supposed to answer the question "where does this quantity belong to?", i.e. what is the theory from which this quantity emerges. It is the 'context', in some sense.
This confusion basically reflects the internal confusion of the encoding in Drasil: it is a flat ontology with no indication of what the relation between the pieces are, nor what 'level' a chunk belongs to (concrete, theory, meta-theory, etc). In part, that's good: the MDE layering get crazy really quickly.
What we really need to do is to revisit the intent of all of the chunks, and come up with a better "meta-model" (sigh) for it all. Then we can figure out what a good implementation could be.
What we really need to do is to revisit the intent of all of the chunks, and come up with a better "meta-model" (sigh) for it all. Then we can figure out what a good implementation could be.
Sounds good. I think we've mentioned this enough times that we should just do this soon. Perhaps we can add it to one of our next in-person meetings?
The agenda for this week's meeting is full, but I've added this to our June 6 meeting (#3453) so that we don't forget to start the discussion. I agree that the intent of our chunks keeps coming up in discussions.
