But its complicated

• Theorem: If there is a normal form for a lambda expression, then it is unique

  • There isn’t always a normal form, however

• Theorem: If there is a normal form, then normal-order reduction will get to it

  • Applicative-order reduction might not

•  So, it seems pretty clear that you want to define a functional language in terms of normal-order reductions, right?  

Previous slide

Next slide

Back to first slide

View graphic version