Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!cs.utexas.edu!yale!think!samsung!xylogics!world!bzs
From: bzs@world.std.com (Barry Shein)
Newsgroups: comp.arch
Subject: Re: Portable Unix boxes [was "Killer Micros"]
Message-ID: <1990Mar17.034530.3278@world.std.com>
Date: 17 Mar 90 03:45:30 GMT
References: <1990Mar6.184634.14749@smsc.sony.com> <36230002@hpindwa.HP.COM> <1990Mar8.210126.11669@utzoo.uucp> <1990Mar11.033429.5732@world.std.com> <53464@sgi.sgi.com>
Organization: The World @ Software Tool & Die
Lines: 47
In-Reply-To: rpw3@rigden.wpd.sgi.com's message of 14 Mar 90 02:04:37 GMT


>p.s. The relevance of any of this to comp.arch is that no matter how the
>parts of our chips or systems talk to each other, the channels are of
>finite bandwidth. And no matter how we encode the data in our registers,
>there will be noise introduced in moving that data. So at some level the
>Nyquist & Shannon limits put a cap on (sequential) computation speed.
>
>
>-----
>Rob Warnock, MS-9U/510		rpw3@sgi.com		rpw3@pei.com

Thanks for a good explanation Rob. I got all sorts of responses to
this, most trying to point out the Shannon Limit, some snidely, yes
all, I was aware of it, that's what I was referring to, people used
the Shannon limit to "prove" to me and others that you couldn't get
much more than 3Kb/s out of a voice-grade phone line.

But my point lies in your last phrase, and you're reference to things
like QPSK etc.

Just that it's a bit facile for someone to say "it's impossible to get
sufficient bandwidth out of a radio signal to make it interesting for
networking (no, not begging the question of "interesting".)

If I can send 10Kb/s on an RF channel, and I can get 1536 or so
channels to use then I can send (roughly, yeah yeah, preambles etc)
10K ethernet-equivalent packets per second, about 50% faster than the
theoretical limit of an ethernet which I'd call "interesting".

All I need is the ability to mux/demux 1536 RF channels (I believe the
speed is quite swallowable), is that utterly unfeasible? I don't think
so, might take some creativity with the spectrum to let a lot of
people do that. Isn't this the kind of thing the C3I spooks do all the
time? Maybe we can't talk about this.

The obvious thing is to use a few dozen more channels and do ethernet
style addressing in the chips and mux many people onto the same
channel band. Then the FCC probably only has to approve a few ranges
or some such thing per metropolitan area. Eavesdropping of course begs
for encryption, so don't start that.

Anyhow, never say never. I guess.
-- 
        -Barry Shein

Software Tool & Die    | {xylogics,uunet}!world!bzs | bzs@world.std.com
Purveyors to the Trade | Voice: 617-739-0202        | Login: 617-739-WRLD