Re: FFT test with few kbits

From: Cristiano (cristiano.pi_at_NSquipo.it)
Date: 03/01/04 

Date: Mon, 01 Mar 2004 13:57:05 GMT

Paul Pires wrote:
> Ernst Lippe <ernstl-at-planet-dot-nl@ignore.this> wrote in message
> news:4042457a$0$1154$ba620dc5@nova.planet.nl...
>> On Fri, 27 Feb 2004 21:00:14 +0000, Paul Pires wrote:
>>
>>
>>> Take a moment to look at the common sense. Say this test
>>> is deployed, say it routinely passes good random streams.
>>
>> But what evidence do you require before you would say that good
>> random streams routinely pass this test?
>>
>> This would be easy, if we had a proof that Christiano's statistic
>> (asymptotically) follows the KS-distribution, but I hope that you can
>> agree, that we currently don't have such a proof.
>
> I can agree with that.

You agree? Do you want the proof that my statistic follows the
KS-distribution? What is the KS-distribution? :-)

> What this feild needs is a real, working, rational
> program of quality assurance for tests. Cristiano has shown a gift in
> this area although from time to time he gets a little "over animated".

I'm not over animated. I hate to see guys who insist in their wrong claims
beyond any reasonable doubt.

Ersnt showed this behaviour also in the thread in which I talked about the
error in the overlap sum test in diehard (thread "[Diehard] Overlap sum
test"). He claimed that if Marsaglia has wrote the test, *surely* I was
wrong somewhere. But when Marsaglia said that the implementation was wrong,
then Ernst fell silent.
My thought is that Ernst is posting his inconsistent messages because my
name is just Christiano instead of George.

Cristiano