There appears to be a connection at the hip between the names of the URLs below a Thing's root URL, and entries in JSON documents served from URLs further up. This connection should be generalized and documented.

Suggestion: define all the "atomic" interactions at lower-level URLs first. Then, look whether efficiencies can be gained by packaging several of those "atomic" interactions into a compound one at a higher-level URL. If so, use the documented generic scheme, e.g. "if you perform HTTP GET on each of ROOT/foo/1 to ROOT/foo/27", and concatenate the results into a JSON array, it will produce the same result as if you HTTP GET on ROOT, JSON entry "foo".