On 6/13/07, Michael Casadevall <[EMAIL PROTECTED]> wrote:
I'm looking at Linux and some of the BSD's implementations of random using kernel data. Linux uses a few complex math equations to make the data even more random and various sources. I'll use the equations from Linux since I'm a total math retard to write my own set, and pull random bits from the same source Linux does. I'll probably have a working prototype in a few days of the kernel code; then it would just be necessary to stick a translator on top of that (which will have the option to get entropy from sources other then the kernel or in addition), and we'll have a proper /dev/random.
Hum, the code used by linux and BSD is to provide an entire PNRG solution. One possible solution to the GNU/Hurd project would be to let the kernel only deal with the entropy gathering itself, and thus leaving all of the difficult math to a translator. One of the source code examples on the wiki does exactly this, Storing the entropy in a pool, and having a translator on top using the entropy gathered in a translator. Anyone want to comment on what we should use to judge the quality of
the entropy, and what is going to be needed to get these patches to migrate upstream into both Hurd and Mach Michael
I don't know how to test the quality of the raw entropy itself, but there are a lot of free tools for statistical analyses of PRNG out there, such as the DieHarder[1] Hope I was of any help :) [1] http://www.phy.duke.edu/~rgb/General/rand_rate.php
_______________________________________________ Bug-hurd mailing list Bug-hurd@gnu.org http://lists.gnu.org/mailman/listinfo/bug-hurd
