News posts seem to be broken, this is a resent:
----- Original Message -----
From: <jeroen@php.net>
Newsgroups: php.cvs
Sent: Thursday, September 06, 2001 12:03 AM
Subject: Re: [PHP-CVS] cvs: php4 /ext/standard rand_mt.c rand_sys.c
>> "Andrei Zmievski" <andrei@ispi.net> wrote in message> news:20010905164801.G3305@ispi.net...> > On Wed, 05 Sep 2001, Jeroen van Wolffelaar wrote:> > > Just look at the algorithm and you'll see at once it is bogus. Only in> the> > > special case that one single key is requested it is correct.> > > Proving it's bogus is trivial, so I'll leave that as an exercise to
the
> > > reader (I'm serious).>> There are about 2**31 possibilities for randval. So already when picking
13
> elements out of an array, (which can be done in 13! ways), some> possibilities are not even possible!>> Furthermore, it makes simply no sense, just look at the key point, every> time r/R, where r and R decrease... that has a really high preference for> 'neat' distributions.>> --Jeroen>>>> > A little less patronizing, please.>> I'm very sorry, but this again is an example of simply test things out> without reasoning/thinking/proving. Which is IMO very bad.> If you simply think about the algorithm, I'm sure you would have seen> immediately what was wrong. I must say I didn't do any testing at all with> that function, but that's not necessary!>>