From our perspective, that’s just an implementation detail that’s up to you and other developers.
I was only meaning about CSL style schema and instance development; what we do in the CSL project proper.
JSON just isn’t appropriate for those critical bits.
And I could imagine some developer wondering this in the future, so thought I’d amend my post.
I think we are in full agreement then. It is just for these caching purposes that I think there is a point in using json.
By way of update, I’ve started to rethink this. See this more recent thread.
This assumes, which is my basic idea there, that we can indeed simplify and generalize the model sufficiently for it to work in non-XML contexts; most notably JSON and YAML.