From: Chuck Kuecker (ckuecker@ckent.org)
Date: Mon Sep 01 2003 - 13:38:25 MDT
Given a network connection - it's easy to put a tap on the line that picks
up every packet passing in both directions. Once a packet is received, it
can be processed for content or routed to a human operator for examination,
or simply saved for later download. The tap adds no significant delay,
since it's not "active" and it affects all packets travelling on that
channel. It's basically a parallel connection to the wires that adds a very
small loading, but no time delay. A fiber optic link is harder to tap -
would require a specialized coupler in series with the fiber or a buffer
that converted fiber to wire and back again, but it's not something really
hard to do, and this method adds only nanoseconds to the transport time.
There are "sniffer" boxes available that simply tie into an unused network
port, and could allow saving every packet on the network. Not difficult
technically - the hard part would be making sense of all those packets, but
it could be done using a supercomputer, or in parallel processors.
Any network connection can be tapped at an ISP, at the dial-up modem or at
the fiber coming into the switch. It takes a pretty fast computer to handle
traffic on a fiber link, but these exist - and since they are only
receiving and don't need to respond, the processor is actually loaded less
than the switch's processor.
I suppose you could put something into the ISP that re-routed only certain
packets via address, or via content, but that seems like a lot more work
and would add the very delays you are supposing. Much easier to tap, listen
to everything, and only "save" what's of interest. If they can read what's
being sent, why should they give anyone a hint that they are doing so?
Chuck Kuecker
At 12:58 09/01/2003 -0500, Extropian Agroforestry Ventures Inc wrote:
>How in a precise physical sense does it do this?
>Is there a particular way it might route messages of varied degrees of
>interest so that a copy can be made. If there is say an extra hop in the
>transfer and a que as it records and sends along the data. If it needs a
>particular path to process/glean then the most efficient routing would be
>reduced to a bottleneck. There is a definite difference in time lags say from
>a send to self within my own ISP and this list.
>
>Given the enormous amount of data packets and the adgenda of TIA et al to be
>thorough there is bound to be detectable evidence of traffic slowdown to
>enable processing?
This archive was generated by hypermail 2.1.5 : Mon Sep 01 2003 - 13:52:34 MDT