On Tue, Apr 13, 2010 at 01:29:59AM -0400, Aaron Sherman wrote:
: PRNGs are often misrepresented as frivolous, but as I'm sure you know from
: your work at JPL, high quality random sequences are much-prized, and any
: language that starts off with some poor assumptions will ultimately pay for
: it.

Actually, my appreciation for good random number generators stems from
my work even before JPL, at System Development Corporation, where
I helped rewrite their discrete event simulator (known as MODLIT).
The guys there did a lot of statistical analysis of the RNG to
make sure it didn't have various subtle artifacts.  In particular,
I worked there with Peter Montgomery, a mathematician with an Erdős
number of 1.  Pity I never published with him.  :)

The fact that early Perl just tended to pass through the system's rand
function had more to do with trying to be a glue language to existing
APIs than it did with trying to provide a good RNG.  (Plus realizing
that anyone in those days who wanted a good RNG tended to write their
own anyway.)  In any case, there weren't many standardized sources of
entropy in those days.  Back then, if I wanted a decently entropic
seed, I'd run a ps that dumped as much of the internal state of the
OS as possible (including various fast-changing internal tables),
and then hash that.  In strength it was probably somewhere between
/dev/random and /dev/urandom.


Reply via email to