lThe
idea of defining the semantics of lambda calculus by reducing first every expression to
normal form (for which
a simple mathematical denotation exists) by a sequence of contractions is
attractive but, unfortunately, does not work as simply as suggested…
The problem is that,
since every contraction step … removes a l, we have deduced a bit hastily that it
decreases the overall number
of ls. We have neglected the possibility for a contraction step actually to add one l, or even more, while it removes another. – Meyer