cvs commit: src/sys/modules/random Makefile src/sys/dev/random harvest.c hash.c hash.h nehemiah.c nehemiah.h probe.c randomdev.c randomdev.h randomdev_soft.c randomdev_soft.h yar

[Gary W. Swearingen]

(I'm sorry if I didn't manange to get this into the appropriate
thread; I'm not subscribed.)

Does a good PRNG (and Yarrow) attempt to generate numbers that look
truly random (albeit with a very long repeat) or does it try to
generate numbers that are better (and thus different) than truly
random numbers, in some ways that make sense for common uses?  For
instance, do good PRNGs give testably-fewer long sequences of zeroes
or ones than you'd get from a truly RNG?  (Some users hope so.)

If there's supposed to be a difference, then would it be useful to
offer a best attempt at truly RNs?  I don't know who needs them,
though, and I have no idea what post processing of the VIA HW RNG
would be needed for that, if any.  Maybe none.

Also in that case, common users should definitely have some post-
processing to remove the long strings of zeros, etc., to make the HW
RNG look more like a good PRNG.  (I wonder if VIA RNG's "string
filter" has something to do with avoiding long strings of zeros or
ones, but I don't care enough to look, since it probably wouldn't be
the best way to handle the problem (?) in FreeBSD.)