Mixin inheritance clarification

[niegrzybkowski]

Closes linkml/linkml#3849

[matentzn]

I think this is correct.

EDIT: withdrew some minor concerns, misread the issue.

[matentzn]

"direct ancestor" is a bit of a contradiction, so better use the word "parent" (I dont mind the redundant "direct"):

Suggested change

	- for any direct (`is_a`) ancestor `a` of `c`, if `a.mixin` is True, then `c.mixin` SHOULD be True
	- for any direct `is_a` parent `a` of `c`, if `a.mixin` is True, then `c.mixin` SHOULD be True

[matentzn]

Suggested change

	- for any direct (`is_a`) ancestor `a` of `s`, if `a.mixin` is True, then `s.mixin` SHOULD be True
	- for any direct `is_a` parent `a` of `s`, if `a.mixin` is True, then `s.mixin` SHOULD be True

[gouttegd]

(Asked to review by @matentzn )

(i) I agree with the assessment in the original ticket that the current phrasing in the spec is wrong, as it would indeed imply that any class that uses a mixin (has a mixins parent) should itself be a mixin, which wouldn’t make any sense.

(ii) I agree with @matentzn ’s objection regarding the use of “direct ancestor” as a way to refer to a “is_a parent”. This is not how most people would understand such a term. In the absence of any established terminology to distinguish between “parent through a is_a relationship” and “parent through a mixins relationship”, better to always be explicit and to always talk about “is_a parent” and “mixins parent”.

So overall, agreeing with the general idea, but requesting the changes suggested by @matentzn.