[117672] in Cypherpunks
Re: Build a better OTP?
daemon@ATHENA.MIT.EDU (Anonymous)
Wed Sep 8 17:10:56 1999
Date: Wed, 8 Sep 1999 22:47:32 +0200 (CEST)
Message-Id: <199909082047.WAA11689@mail.replay.com>
From: Anonymous <nobody@replay.com>
To: cypherpunks@cyberpass.net
Reply-To: Anonymous <nobody@replay.com>
> >> ...this shouldn't be a "Windows vs. Linux" issue at all!
>
> > Right, it's not; it's a binary-API issue. The Linux community
> > will not accept a binary API. They want to see source...
>
> *Sigh*
>
> Let me repeat: "this shouldn't be a "Windows vs. Linux" issue at all!"
>
> My point was that it's bad engineering practice for Intel to design a
> hardware RNG that requires an OS-specific API.
But Intel didn't design a hardware RNG that requires an OS-specific API.
They could make a Linux API without any problem. However it would be
in object code form and would not give direct access to the raw bits
coming out of the hardware, for the many reasons already enumerated in
this thread, one of which is that it allows them flexibility in changing
the design in the future.
Hence your point is irrelevant since Intel did not do what you claim.
> If they (legitimately)
> want to protect their IP, they should implement the whitening in
> microcode, and decouple the RNG entirely from the operating system.
This is absurd. The RNG is not on the CPU chip at all, it is on a bus
controller chip, and has a limited space budget because the chip needs to
perform many other functions. There is absolutely no room for something
like a SHA whitener in microcode. It would make the RNG at least an
order of magnitude larger.
Once again we have an ideologically based objection which ignores the
reality of engineering constraints in the real world.
