From:	Robert Nesius <nesius@******.COM>
Subject:	Re: Fiber Optics [was: Re: FAB Revisited]
Date:	Thu, 14 May 1998 23:06:55 -0700

Hi there,
I'll offer a perspective on how a one-way dataline may work. This might be
too "grounded in reality," but what the heck.

I think one way you could look at one-way dataline is that it's one way
not because of physical restrictions of the medium (the physical layer),
but because of restrictions imposed in higher layers of the protocols that
practically by definition must exist in networking. The protocols
that don't care what the bits they are schlepping around are, because their
job is to just send it on.

Within this context, it is possible to configure nodes on the matrix
responsible for routing and traffic control to not return confirmation
or route information that is not authorized. So the one way data-pipe
is a conceptual construct that is a result of how the matrix is configured
at that particular place. IMHO, jumping into a one way data-pipe
such as this would be like jumping into a sensory deprivation tank.
"Gee, it's nice and dark in here." :) Or, if the decker figured out how
to piggy back his signal on the authorized data stream, perhaps he could
do some damage.

I just don't buy the concept of a truly one-way data-pipe being implemented
in physical layers and used for anything remotely important. One of the
fundamental problems in networking is making sure what was sent was what
was received. At some level, the hardware at the ends of the datapoints
must talk to each other to verify this. If they don't, then the data runs
a risk of being compromised.

In the end, I don't buy the idea of this being implemented in the physical
layer. Perhaps too much reality on part though.

-Rob