Talk:Pseudorandom number generator

About Matsumoto's and Nishimura's Mersenne Twister I can say that it really works almost as Nature's randomness - believe it or not. This algorithm is already fully implemented on VOG a Russian server for board games. I've played there quite vast period of time on Backgammon arena and on Reversi one. The C function int rand (int) is not as good as PRNG based on Mersenne twister - that is for shure. Try it and feel it. Nice you've started a PRNG at last. We really needed this one. I was already thinking to do your job. Can you put some Knuth's work on this field too. About the simpliest PRNG and such.
XJam [2002.03.23] 6 Saturday (0)

> It has a colossal period of 219937-1 iterations, is proven to be equidistributed in 623 dimensions. Ostrich can you please explain a liitle bit more what herein dimensions means exactly? Otherwise everything else is explained as clear as can be.
XJam [2002.03.23] 6 Saturday (1st ed)

For some reason my Delphi/Pascal code is not formated correctly with the <pre> tag. Suggestions? - Jim

Statistical patterns in a sequence may appear in many ways. The dimensions noted are ways of examining the sequence for patterns. See Knuth Semi-numerical algorithms for more detail. Please note that this article needs serious revision, particularly with regard to CSPRNGs. It is not easy to build these and statements that 'one could' have one if one does this or that are dangerous. Someone might rely on them and get into considerable (insecure) trouble as a result. Indeed, many PRNGs are built with block cypher encryption algorithms operating in CTR mode. However, the description of CTR mode here is inadequate. Note that CSPRNGs not only must satisfy certain statistical properties, but also must not be predictable which is another kettle of bits altogether.

Move Cryptography Material

I see that most of the material in Cryptographically secure pseudorandom number generators is repeated on its own page. Can we remove or summarize it here?

Sander123 13:13, 21 Jul 2004 (UTC)

Sander, The biggest use of RNGs is in crypto by now and so some (all?) of this coverage belongs here. Some RNGs suitable for other uses are not suitable for crypto use, so some attempt to bring out the different requirements is apposite.

As for duplication, much the same can be said for coverage of the material from a crypto perspective, so a redirect of one title to the other will probably cause difficulties. I too feel an urge to minimize duplicated content, but in this case have managed to restrain myself. I suspect we all should. ww 14:53, 21 Jul 2004 (UTC)

ww: "The biggest use of RNGs is in crypto" — are you sure? — Matt 23:16, 21 Jul 2004 (UTC)

A general rule of thumb is to have only a summary paragraph about a subtopic, and include a prominent link to the subtopic article; I think we do this here, although the subtopic article (CSPRNG) is not as complete as it probably should be, and arguably the summary paragraph in PRNG is slightly too long as well. — Matt 23:16, 21 Jul 2004 (UTC)

I agree with Matt. The problem that I see is that repeating the material makes maintanace difficult. In fact I came across this because I wanted to add some material. And there is a lot of material that I want to add here--Make the requirements more explicit, say some more about the standards availlable, list designs that were broken and the problems they caused, etc, etc.

Eventually this could be quite a large topic. Should I add all that to the main page? I propose that I make sure that everthing on the main page is included in crypto page, and then summarize the main page. Everybody agree? Sander123 08:50, 23 Jul 2004 (UTC)

Seems like a sound refactoring to me, thanks. — Matt 17:00, 23 Jul 2004 (UTC)