This is cool. It also raises the possibility of adding tail call optimisation to JavaScript interpreters.

Anyone know if that would be feasible?