random seeds

    May 10 2006 | 11:28 am
    Hi randomisers...
    I know that random and urn can be seeded from a number dependent on the time since system start-up (seed 0) but I'm working on something that will start-up at the same time every day, and various parts of the program will be triggered at fixed times. This suggests that theoretically I could get the same sequence of events happening every day. << or is this extremely unlikely?>>
    To avoid this i just thought of generating the seed from sampling a piece of incoming audio (microphone) - remembering that some computer (atari1040ST?) used to provide random numbers by sampling some noisy piece of circuitry -
    is there a simpler, cooler way to do this? do anyones externals generate random seeds in a different way?
    - i looked in the archives but people seemed to be more interested in repeating a random sequence of events!
    groeten, justin.

    • May 10 2006 | 1:40 pm
      you could use a prime number generatorr... i.e. tap.prime
      or use the output of the object itself multiplied by the cpuclock ?
    • May 10 2006 | 5:59 pm
      Hi Brian,
      I was just about to say that I can't see why either of your suggestions would make it more random. but now it suddenly dawns on me while typing: you're right about the system clock because of course the date component is different every day.
      But why Prime?
      thanks, justin.
    • May 11 2006 | 7:07 am
    • May 11 2006 | 9:26 am
      afaik computers report time since startup in ticks (1/60th of a second), so that would be unlikely, or at least unlikely to where the sequence might repeat once or twice a year, which you would be highly unlikely to notice. Personally I use the julian date&time.
    • May 11 2006 | 9:51 am
      yeah, once or twice a year no-one is going to notice! Right now i'm making a seed by mangling the current date into a useful number. justin.
    • May 11 2006 | 10:00 am
      maybe someone should modify the date + time object to accept the commands julian / orthodox / chinese / jewish / islamic / mayan / etc... in the interests of temporal diversity. ok, someone has probably done it already!
    • May 12 2006 | 4:24 am
      On around May 10, 2006, at 13:28, justin bennett said something like: >is there a simpler, cooler way to do this? do anyones >externals generate random seeds in a different way?
      The entire suite of Litter Power RNGs base their seed on multiple items drawn from the OS that will not repeat themselves. Ever. They also use an industrial-strength random number generator, rather than the Linear Congruence generator used by random and urn.
      If you're seriously interested in random numbers, you owe it to yourself to take a look at Litter Power. The Starter Pack is free. URI below.
      Best, Peter
      -------------- http://www.bek.no/~pcastine/Litter/ -------------- Peter Castine (HTML doesn't like the rest of this .signature)