You may very well be right. I did notice that in the article that Roy Johnson linked on the Haskell implementation the author spent longer than I would have expected on the discussion of the random numbers, which gave me the impression that the generation of random numbers has a conceptually different place, or different standing if you will, in the Haskell world than I'm used to from other programming languages.

But to be honest, I barely understand what this requirement that the shuffle be "functional" is, as is probably plainly obvious from my various stabs at it.