Explain why you want to implement infer_shape.

doc/extending/op.txt

@@ -153,6 +153,10 @@ following methods:
    The function should return a list with one tuple for each output.
    Each tuple should contain the corresponding output's computed shape.
 
+   Implementing this method will allow Theano to compute the output's
+   shape without computing the output itself, potentially sparing you
+   a costly recomputation.
+
 .. function:: make_thunk(node, storage_map, compute_map, no_recycling)
 
    TODO