1996-01-08 - A /dev/random standard is need.

Header Data

From: wlkngowl@unix.asb.com (Mutatis Mutantdis)
To: N/A
Message Hash: 13031d20de01a3c41c2f3d2cabcdaaae279c8b8a4469e3b03725ddf6c4d3f16e
Message ID: <DKt3sx.IoA@news2.new-york.net>
Reply To: N/A
UTC Datetime: 1996-01-08 02:02:58 UTC
Raw Date: Mon, 8 Jan 1996 10:02:58 +0800

Raw message

From: wlkngowl@unix.asb.com (Mutatis Mutantdis)
Date: Mon, 8 Jan 1996 10:02:58 +0800
Subject: A /dev/random standard is need.
Message-ID: <DKt3sx.IoA@news2.new-york.net>
MIME-Version: 1.0
Content-Type: text/plain



I'm revising the DOS NOISE.SYS driver currently.  In writing the
documentation I am discussing the advantages of such a driver (part of
the logic behind writing it).

Mainly, a hardware (and to some extent) operating system independent
means is needed for generating random numbers.  It seems to me that a
device driver (at least for DOS, Unix and maybe Amiga or Atari) is the
best way to do this.  If one has special chips or diodes for
generating randomness, a device driver which reads from them can be
used. If one lacks such equipment, something like NOISE.SYS or
random.c for Linux, or Noiz (which I have not yet look at) can be
used.

Even then, random.c defines two devices, random (which only returns as
many bits as there are fresh bits in the entropy pool) and urandom
(which keeps hashing the bits and will return as many as requested.)

NOISE.SYS defines only random, which behaves more like urandom above.

If there is a standard, it will make it easier to use special hardware
since software which reads from a random device can access it.

Perhaps a kind of standard should be discussed and created so that
cross-platform development is made much easier, and so that features
and capabilities can be worked out.

Does anybody else see a need for this?


--Rob







Thread