So are we all.
I’m heedful though of the corpses of 20 SE clones already, and here you want to out-do SE itself eventually.
It’s that threaded conversations are kind of tree or graph.
And documents are trees too, nested subsections, maybe you want different rules (tags, privileges, commentary) for some section (and subsections) of a book and not others.
Looking back I guess see this and especially e.g. this.
Plus Monica’s saying, “When SE did Doc they had to make a whole new thing” though I don’t know about that.
Maybe we want to be working with trees and structured documents, and relational is better for implementing mere scrapbooks.
The example in Celko’s book I referenced, of “hierarchical” information, was of “parts with sub-parts”.
So SE now is all about 1-to-n relations of different kinds – not about different layers of the same kind of information (tags with subtags, sections with subsections) – and the lack of hierarchy makes the data a big flat unstructured mess (which you can search with a search tool but not organise).
I think that’s fundamental and an artefact or constraint of the database implementation, or am I wrong?
See also ontology I guess (though I never studied comp.sci.).