[Haskell-cafe] lazy evaluation is not complete

A simplied version of Example 5-16 in Manna's classical book "Mathematical Theory of Computation": foo x = if x == 0 then 0 else foo (x-1)*foo (x+1) If run with ghci, foo 5 does not terminate, i.e., Haskell does not look for all outermost redices in parallel. Why? For efficiency reasons? It's a pity because a parallel-outermost strategy would be complete.