This is another attempt to solve the problems that inspired this proposal which now seems to me to be too heavy. This version presumes that most identifiers bypass any cell. As in Algol68, Ocaml and several other languages one must declare variables differently if you need to mutate them. Unlike Algol68 but like Ocaml references to the value requires different syntax. Most of my Scheme programs would be valid as written in this language variation for I seldom use set!.
This version admits references to the language ontology. Perhaps it is not different from Rees’ ‘cells’.