> Joins and source tables are automatically resolved in Trilogy. You won't ever explicitly specify one in your query; you're declaring what you want, not how to get it. The responsibility of how to get it is delegated to the semantic model.
I don't fully understand how the semantic model is created (is this documented anywhere?), but I don't think I would enjoy a hard separation between the query layer and the semantic layer regardless.
I would prefer a continuum of "how much indirection do I really want" to be available to the user. My own exploration of this topic is https://docs.racket-lang.org/plisqin/index.html and you can compare section 2 (Using define-schema) to section 7.2 (Plisqin Desugared) if you want to know what I mean about a continuum.
Unrelated, the article we are commenting on has inspired me such that I think I have an answer to the big type system questions that eluded me when I decided to put Plisqin on the shelf. Maybe time to pick it up again... but probably not.