> I'm having trouble picturing what such restrictions would look like. [...] there's no sane way to restrict cdr. I guess you could question its sanity, but th

> I'm having trouble picturing what such restrictions would look like. [...] there's no sane way to restrict cdr. I guess you could question its sanity, but the obvious way would be to say that any object (cons cell) produced by make-closure or (successfully) passed to apply is/becomes a "closure object", and car/cdr will either fail at runtime or deoptimise it into its 'official' representation.