[Python-ideas] Should our default random number generator be secure?

[Steven D'Aprano]

On Mon, Sep 14, 2015 at 10:19:09PM +0100, Robert Kern wrote:

> The requirement for a good PRNG for simulation work is that it be *well* 
> distributed in reasonable dimensions, not that it be *exactly* 
> equidistributed for some k. And well-distributedness is exactly what is 
> tested in TestU01. It is essentially a collection of simulations designed 
> to expose known statistical flaws in PRNGs. So to your earlier question as 
> to which is more damning, failing TestU01 or not being perfectly 623-dim 
> equidistributed, failing TestU01 is.

I'm confused here. Isn't "well-distributed" a less-strict test than 
"exactly equidistributed"? MT is (almost) exactly k-equidistributed up 
to k = 623, correct? So how does it fail the "well-distributed" test?


-- 
Steve