Random Number Generation -----> Hardware or Software?

Evidence, please.

Because the universe is finite, and thus the PRNG cannot increase its workspace without bounds. Sorry, you are being too imaginative.

--
Guy Macon

Wrong. Most Linux programming languages use the output of /dev/random or /dev/urandom to seed the language's RNG, and /dev/(u)random gets random bits from physical sources.

No, they do NOT "have to." The period can be much longer than the age of the universe.

Happens all the time[1] when using RC4 as a PRNG.

[1] (Roughly as often as it happens when using dice or coin flips...)

Wrong again.

Most PCs are *not* deterministic. The turbulance of the air inside the hard drive causes variations in access time, for example. Modern operating systems take these variations and apply a strong cryptographic hash to generate nondeterministic unpredictable numbers.

If you had actually tried this on a cryptograpic PRNG you would already know that it won't work.

If you had actually tried this on a cryptograpic PRNG you would already know that it won't work.

I strongly suggest that you do some research before expressing any further wrong information.

No problem. Enroll on a serious statistics course, and all will be revealed. I do not, of course, mean Remedial Statistics for the Mathematically Impaired.

You clearly haven't looked at the published universal tests. All of them need an unbounded state space. Oh, sorry, I forgot that you haven't been on the statistics course yet.

Regards, Nick Maclaren.

Riiiight. The best cryptography experts in the world say that a cryptographically strong PRNG is indistinguishable from random data, the best known software for identifying bias (DIEHARD) cannot find bias in cryptographically strong PRNGs, yet I am supposed to believe that this unnamed method is taught in statistics courses. Suuuure it is.

Look here for evidence that you are wrong:

And this allows an unbounded state space to fit inside a bounded universe - how?

The boundedness of the universe is not a settled question.

Regards Emil

True, and anyway it's kind of irrelevant, isn't it? The set of positive integers, for example, is an infinite set whether or not the universe is infinite.

Ed

"Eric Smith" wrote

I should amend the above to: "Prng's pass randomness checks with

100% flying colors and that's why they fail. They cheat on the test."

Yes, it is theoretically possible to create a prng with some arbitrarily long repeat cycle. And one can postulate that in some arbitrarily long lived universe the pattern will repeat.

That is not the point. The pattern is not random. It is deterministic. If the generating method is known then each number coming out is known with 100% certainty. The only thing random about such a generator is the algorithm. The seed at which it starts is irrelevant to the discussion: the sequence is circular - any starting point on the circle is equivalent to any other.

It is like reading a book (it is exactly like reading a book), though one may not know the next word, there is only _one_ possibility. There is _no_ uncertainty. That the observer is ignorant does not make the sequence random.

A prng is nothing more than a _counter_. The old TMS1000 micro used a prng for the program counter because it could be made with less silicon than a 'real' counter. Debug was a gas ...

A zener diode and a comparitor will produce a sequence that can not be predicted. It being a real implementation it will have bias but in the limit it becomes impossible to tell bias from 1/f noise. The next generator built will have a different bias, etc. etc.

Good Lord. Get thee to a palmist and have a seance with von Neumann and Turing and ... Far better minds than those here have declared PRNG's to be mirage.

As the saying goes "The mind of God is unknowable", and I would add "even to God".

--
Nicholas O. Lindan, Cleveland, Ohio
Consulting Engineer:  Electronics; Informatics; Photonics.
To reply, remove spaces: n o lindan at ix  . netcom . com

I would counter that you are being maybe too imaginative. When growth to infinity is allowed the arguments get silly.

I didn't know they were mutually exclusive. For the record I am a self excommunicated Unitarian.

The argument is free-will Vs predestination. As no one has been able to settle the issue it seems that free will wins. Unless you are of a particularly dour sect of Scots Presbyterian.

Any attempt at predicting reality fails. Popper would say we have to take reality as random until proven otherwise.

Ha Ha! Have you tried it?

I can only recommend doing so. A real learning exercise. A 15V zener, comparator and a parallel port are all you need. In theory one should be able to remove any 1/0 bias with an algorithm, in reality ...

Very, very true. Wait long enough and it will fail and start producing real noise.

If it has defects from the beginning then it isn't a prng, is it?

--
Nicholas O. Lindan, Cleveland, Ohio
Consulting Engineer:  Electronics; Informatics; Photonics.
To reply, remove spaces: n o lindan at ix  . netcom . com

It is even possible to generate a PRNG that does not repeat. It may stop working due to memory limits at some point, but it does not need to repeat. Like a algorithm that produces all digits of pi, it will be fully deterministic, though.

Or the thermal noise of a resistor. If you use an auto-zero amplifier, the bias goes away, but the noise level depends on temperature. Add an automatic gain control (which takes the noise amplitude only), and you get rid of that influence, too.

--
Bernd Paysan
"If you want it done right, you have to do it yourself"
http://www.jwdt.com/~paysan/

"Guy Macon" wrote

These the same crypto experts who bring us all these crackable codes. A prng is just one more code. It is crackable just like any other. By _definition_.

Philosophy, literature, art, religion ... stats hasn't caught up yet. In my mind PLA&R kick over rocks for stats to come and explain at some later date.

Funny. None of them say prngs are really-truly random. They say you can get 'good enough'.

And I didn't know Google was a particularly good source for the truth:

And

And if you liked the timecube:

And I can't believe I agree with Macon. Aw, it's just a random event.

--
Nicholas O. Lindan, Cleveland, Ohio
Consulting Engineer:  Electronics; Informatics; Photonics.
To reply, remove spaces: n o lindan at ix  . netcom . com
psst.. want to buy an f-stop timer? nolindan.com/da/fstop/

In order to prove that reduction to absurdity is not a valid method of proof, I have set a machine (down in the basement) to discovering and displaying the largest prime. It does this quite simply, by displaying all numbers that are not the largest prime. It is making steady progress.

For efficiency reasons it does its displaying in hex. So far I can confidently assert that the largest prime is not less or equal to:

0xffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff

The result will eventually require confirmation, because the ECC memory cannot correct multiple bit errors, in fact it cannot even detect all 3 bit errors. In case someone wants to do this in parallel, rather than after I publish the end result, suitable C code follows:

i = 0; largestprime = 0; while (i >= largestprime) { if (i && prime(i)) largestprime = i; printf("0x%x\n", i); if (!++i) break; }

and there are plenty of published methods of evaluating prime(i).

Note: the C system must have a suitable value of INT_MAX, otherwise the above code can exhibit undefined behaviour.

--
"If you want to post a followup via groups.google.com, don't use
 the broken "Reply" link at the bottom of the article.  Click on 
 "show options" at the top of the article, then click on the 
 "Reply" at the bottom of the article headers." - Keith Thompson

Q: Is a Pseudo-Random Number Generator's output functionally equivalent to a random sequence?

A: PRNG: The next number in the sequence is knowable with sufficient knowledge. Not being able to predict the number is due to ignorance and has nothing to do with 'random'.

Random: No amount of knowledge will allow prediction of the next number.

--
Nicholas O. Lindan, Cleveland, Ohio
Consulting Engineer:  Electronics; Informatics; Photonics.
To reply, remove spaces: n o lindan at ix  . netcom . com

So what? The universe can be much longer than the period.

--
Nicholas O. Lindan, Cleveland, Ohio
Consulting Engineer:  Electronics; Informatics; Photonics.
To reply, remove spaces: n o lindan at ix  . netcom . com

"Robert Finch" wrote

Douglas Adams' "Infinite Improbability Drive"?

Good point you bring up: does a PRNG/Pi create entropy?

--
Nicholas O. Lindan, Cleveland, Ohio
Consulting Engineer:  Electronics; Informatics; Photonics.
To reply, remove spaces: n o lindan at ix  . netcom . com

On Intel's motherboard/chipset RNG:

-- Nicholas O. Lindan, Cleveland, Ohio Consulting Engineer: Electronics; Informatics; Photonics. To reply, remove spaces: n o lindan at ix . netcom . com

"Guy Macon" wrote

I think that just about ties it up for me.

Guy, baby, put me in your kill file and lead a happier life.

--
Nicholas O. Lindan, Cleveland, Ohio
Consulting Engineer:  Electronics; Informatics; Photonics.
To reply, remove spaces: n o lindan at ix  . netcom . com