Weekend Gedankenexperiment - The Kill Switch
dhc2 at dcrocker.net
Sun Feb 6 13:09:05 CST 2011
On 2/6/2011 10:47 AM, Barry Shein wrote:
> If you focus it down very sharply like this:
> DARPA specified (or, perhaps, the project was sold to DARPA with
> a promise...) that the network being designed in the late 1960s
> should be resistant to a nuclear attack.
> That's probably an urban legend, who knows, it's probably not all that
> But was it observed over and over from the early on that a packet
> network, versus the then predominant technology of virtual (or even
> real) circuit networks, would be resistant to damage of all sorts?
Sorry, but I think the technical implications of a goal to survive 'hostile
battlefield conditions' versus 'nuclear attack' are (small pun) massively
different. Hence I think the actual language used matters.
And the fact that the common language around the net during the '70s was the
former and not the latter matters. Which is why it would be helpful to get some
credible documentation about use of the latter.
I'd expect the major difference in the two terms is the scale of the outage. A
few square miles, versus possibly thousands.
To that end, I remember an anecdote about van Jacobson from the 1989 quake in
California that might provide some insight about a large-scale outage:
He was living in Berkeley but was visiting Stanford when the quake hit and he
wanted to check that his girlfriend was safe. Of course, the phone didn't work.
Out of sheer frustration and the need to do /something/ he sent her an email.
He got a response within a few minutes.
Surprised that the net was still working (and working quite well), he did a
traceroute from the Stanford system to the one his girlfriend was using.
Not surprisingly, the path did not cross the San Francisco Bay, as it normally
would have. Instead it went down to Los Angeles, across the southern US, up the
East Coast and back across the Northern U.S.
Although the outage was fairly small-scale, the scale of the re-routing suggests
that a larger, 'regional' outage from something like a nuclear event would adapt
readily. (We can ignore the additional question of EMP effects, since that only
affects the scale of the outage.) And, of course, there have been other test
cases since then...
 This is anecdotal; I've never confirmed the story with him.
 That does not automatically indicate a system failure, given the switch to
an emergency mode for the phone system that restricts access during major events
 Van created traceroute. <http://en.wikipedia.org/wiki/Traceroute>
More information about the NANOG