[lkml]   [1996]   [May]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: CONFIG_RANDOM option for 1.99.2
On Tue, 28 May 1996, Martin.Dalecki wrote:

> > This is difficult, because once the information used by the random driver
> > is filtered to the application level, significant amounts of randomness
> > are lost, because of timing considerations. Real randomness generated
> > (esp. on a multiuser OS) by timing numbers can't be filtered. The only
> > way to ensure this is to build it into the OS.
> >
> Hey do You know howsimple they are doing it in X11R6 and they magic cookies?
> They are simply checksumming /dev/mem (in fact using the MD5 thereafter)!
> And now please tell me why this doesn't involve more "entropy" from
> hardware then /dev/random????? Every interrupt will leave the memmory in
> some different state!!!

Because the state the memory will be in is predictable; the seek times on
the disk drive (in milliseconds) is unpredictable by any means (outside
of a given range). In other words, a computer will only produce
deterministic results (ie the state of memory); other hardware is
necessary to produce nondeterministic results. A "different" state
doesn't mean anything if the next state is predictable, which, in the
case of memory, it is. Not to an outsider, but that's not really relevant
to the question of "true" randomness.

Jon Lasser (410)532-7138 - Obscenity is a crutch for inarticulate motherfuckers.
Finger for PGP key (1024/EC001E4D) - Fuck the CDA.

 \ /
  Last update: 2005-03-22 13:37    [W:0.052 / U:1.588 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site