Re: [9fans] Interesting research possibility?

Re: [9fans] Interesting research possibility?

[presotto]

[-- Attachment #1: Type: text/plain, Size: 364 bytes --]

That's essentially how we use /dev/random in the security library.  I wouldn't mind
replacing the current seeding with something a little more trustworthy.  I'm afraid
that beating the cpu clock against the interrupt one will eventually fail because
they'll end up being the same crystal on some system.  On some architectures, like our
SGI's they already are.

[-- Attachment #2: Type: message/rfc822, Size: 1760 bytes --]

From: Dan Cross <cross@math.psu.edu>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] Interesting research possibility?
Date: Tue, 25 Jun 2002 15:59:46 -0400
Message-ID: <200206251959.PAA19776@math.psu.edu>

> I had a few that I used to run at it every now and then; self correlation,
> and a random quadrant walk.  Haven't in a while.  I'll fix them up and stick
> them out there.  We could always use more.

What about doing Yarrow and seeding it with random thermal noise?

	- Dan C.

http://www.counterpane.com/yarrow-notes.html

Re: [9fans] Interesting research possibility?

[Dan Cross]

> I had a few that I used to run at it every now and then; self correlation,
> and a random quadrant walk.  Haven't in a while.  I'll fix them up and stick
> them out there.  We could always use more.

What about doing Yarrow and seeding it with random thermal noise?

	- Dan C.

http://www.counterpane.com/yarrow-notes.html

Re: [9fans] Interesting research possibility?

[presotto]

I had a few that I used to run at it every now and then; self correlation,
and a random quadrant walk.  Haven't in a while.  I'll fix them up and stick
them out there.  We could always use more.

Re: [9fans] Interesting research possibility?

[rog]

> Also, it looks like T23's have a reasonable thermal noise
> hardware random number generator that I'll try to use.  I
> haven't run it against any of the RNG tests yet.

that doesn't include T22's too, by any chance?

i think random number generation should be built in
to more hardware; generating random numbers for
crypto keys seems slow in software... and is it really dependable?

has anyone run a reasonable set of tests on plan 9's /dev/random
under different run-time conditions?

  rog.

Re: [9fans] Interesting research possibility?

[presotto]

On Tue Jun 25 11:14:24 EDT 2002, rog@vitanuova.com wrote:
> > What's left on the screen after the galaxies collide is a *very*
> > random assortment of colors.  Suppose we take a quadrant sample of
> > that and generate a number out of it.  Would this work as a very
> > simple random number generator?
>
> sounds like a fairly elaborate random number generator to me!
>
> however, i guess this isn't so far removed from Wolfram using a
> cellular automaton as the basis for the pseudo random number
> generation in mathematica...
>
>   rog.
>

I still prefer the lava lamp.

Also, it looks like T23's have a reasonable thermal noise
hardware random number generator that I'll try to use.  I
haven't run it against any of the RNG tests yet.

Re: [9fans] Interesting research possibility?

[ozan s. yigit]

| 				... A friend suggested
| watching patters of raindrops as an input.  That is, until
| we figure out how to calculate that.

take a look at SGI's lava lamp random number generator for some ideas.
the description (if memory serves, converting images of a lavalamp into
numbers) should be in the SGI patent. also watch for www.lavarand.org for
the next version. 

[it occurs to me that one can snarf images coming from north american
highway cams. oh, i better go patent that, trademark "trafrand" and write
a journal paper about it. :-P]

oz
---
the only zen you find on the tops of mountains is the zen you bring up there.
                                                         -- robert m. pirsig

Re: [9fans] Interesting research possibility?

[rog]

> What's left on the screen after the galaxies collide is a *very*
> random assortment of colors.  Suppose we take a quadrant sample of
> that and generate a number out of it.  Would this work as a very
> simple random number generator?

sounds like a fairly elaborate random number generator to me!

however, i guess this isn't so far removed from Wolfram using a
cellular automaton as the basis for the pseudo random number
generation in mathematica...

  rog.

Re: [9fans] Interesting research possibility?

[Sam]

Right, I should have seen the pseudo-ness of this approach.  I'm
thinking out loud here and those who've thought through this
will likely say it's obvious, but it seems that the only true
random generators are those whose inputs are based on something
in nature we have yet to find a formula for.  A friend suggested
watching patters of raindrops as an input.  That is, until
we figure out how to calculate that.

Don't smirk, it wasn't *too* long ago that people would
have said, "why not just use the position of the planet Mars
as your input.  He's all over the sky."

Cheers,

Sam

On Tue, 25 Jun 2002 rog@vitanuova.com wrote:

> > What's left on the screen after the galaxies collide is a *very*
> > random assortment of colors.  Suppose we take a quadrant sample of
> > that and generate a number out of it.  Would this work as a very
> > simple random number generator?
>
> sounds like a fairly elaborate random number generator to me!
>
> however, i guess this isn't so far removed from Wolfram using a
> cellular automaton as the basis for the pseudo random number
> generation in mathematica...
>
>   rog.
>

Re: [9fans] Interesting research possibility?

[Chris Hollis-Locke]

Sam wrote...
>What's left on the screen after the galaxies collide is a *very*
>random assortment of colors.  Suppose we take a quadrant sample of
>that and generate a number out of it.  Would this work as a very
>simple random number generator?

[I presume that] the galaxy is seeded using random numbers.
So it would be an _elaborate_  pseudo-random number generator.
It would generate no more (true) randomness than the
original input.

Re: [9fans] Interesting research possibility?

[andrey mirtchovski]

the original galaxy code coloured each galaxy differently, i had it
(for opengl) with colors based on galaxy of origin and speed, but it didn't
look as nice as i expected.

if you're interested i can suggest looking at the xscreensaver code (the
original) for coloring ideas.

i admit, i haven't seen your modifications yet, so i can't comment on them --
for all i know they're probably better than the xscreensaver ones :)

andrey

On Mon, 24 Jun 2002, Sam wrote:

> Sorry I'm so chatty today; I woke up this morning full of fire.
>
> Take the new galaxy.c and change 'effects' line around 351 to the second
> option.  Now run it and select the 'Sparkle' option.  Some thoughts ...
>
> What's left on the screen after the galaxies collide is a *very*
> random assortment of colors.  Suppose we take a quadrant sample of
> that and generate a number out of it.  Would this work as a very
> simple random number generator?
>
> Sam
>

[9fans] Interesting research possibility?

[Sam]

Sorry I'm so chatty today; I woke up this morning full of fire.

Take the new galaxy.c and change 'effects' line around 351 to the second
option.  Now run it and select the 'Sparkle' option.  Some thoughts ...

What's left on the screen after the galaxies collide is a *very*
random assortment of colors.  Suppose we take a quadrant sample of
that and generate a number out of it.  Would this work as a very
simple random number generator?

Sam