While MRMA does try to get a full seed array from random
sources, it is not alway guaranteed to do so. If a full
seed is not acquired, a warning is issued, and the values in
the seed are repeated to fill up the PRNG. In such a case,
get_seed returns a full array with some duplicate values.
For instance, in the case when no sources available, MRMA
will use the PID and time as a partial seed, and then repeat
them over and over until the PRNG is fully seeded. get_seed
will then return (PID, time, PID, time, PID, time, ...).
This phenonomenon accounts for the repeated values in your
table.
For this reason, I don't recommend saving seeds as a source
of random numbers.
That being said, you can ward against partial seeds by
trapping warnings when doing srand, and then skipping the
seed:
use Math::Random::MT::Auto ':!auto';
my @WARN;
$SIG{'__WARN__'} = sub { push(@WARN, @_); };
my $prng = Math::Random::MT::Auto();
for (1..50) { # This will exhaust your daily quota
$prng->srand('random_org');
if (@WARN) {
print('Skipping seed: ', @WARN);
undef(@WARN);
} else {
my @seed = $prng->get_seed();
# Store seed in database
}
sleep(10);
}
You also asked about using /dev/urandom. This device draws
from /dev/random until such time (if ever) that /dev/random
is exhausted. Then it falls back to a PRNG to fill the rest
of the request. So using /dev/urandom is a good idea if
you're not using it too frequently.
If you want to guarantee a fully random seed, however, then
use /dev/random. If you need a full seed, then trap
warnings as per the above and using:
$prng->srand('/dev/random/');
Is MRMA better than rand() even if using just on 32-bit
seed? Yes, because MRMA can produce a longer sequence
before repeating. (I think I give some idea on this in the
POD, but maybe not.)
