I think the key thing we should have learned from LTCM is that it's difficult to determine which risks are independent, which are not, and what sort of feedback the dependent ones will create for one another.