+
+3. In the previous homework, one of the techniques for controlling
+evaluation order was wrapping expressions in a `let`: `let x = blah in
+foo`, you could be sure that `blah` would be evaluated by the time the
+interpreter considered `foo` (unless you did some fancy footwork with
+thunks). That suggests the following way to try to arrive at eager
+evaluation in our Haskell evaluator for CL:
+
+ reduce4 t = case t of
+ I -> I
+ K -> K
+ S -> S
+ FA a b ->
+ let b' = reduce4 b in
+ let a' = reduce4 a in
+ let t' = FA a' b' in
+ if (is_redex t') then reduce4 (reduce_one_step t')
+ else t'
+
+Will this work? That is, will `reduce4 (FA (FA K I) skomega)` go into
+an infinite loop? Run the code to find out, if you must, but write
+down your guess (and your rationale) first.
+
+<!-- Doesn't work: infinite loop. -->
+