The August Forbes ASAP quotes me as saying "Metcalfe has become an elder
statesman and now he is doing more harm than good, spreading fear and doubt
while the rest of us solve problems." What I *said* was "Many people think
that Metcalfe has ...". I believe that was the consensus on nanog (meaning
I read a few posts to that effect from nanog participants and repeated
it to Gilder, who reported it as my original quote); does anyone have the
original nanog list message so I can confirm it in my correction letter to
ASAP? Forbes has a history of getting impressive, if not entirely accurate,
wording to put under photo headings.
They also got all sorts of technical things somewhat wrong, like an explanation
of exponential vs. polynomial-time algorithms.
And when the fact-checkers called to check the piece, they quoted from
the article: "Cisco routers are tuned to ignore pings from the route servers".
What I said, of course, was something like "Using packet loss and ping times
from the route servers to Cisco routers at the exchange point proves nothing
in particular except that Ciscos are tuned to ignore pings and other low-
priority (ICMP) traffic when route processing and moving data around must
be performed". They didn't quite fix it in the final version. But maybe
we can get Cisco to actually ignore all ICMP packets from source addresses
of the currently-existing-as-of-the-IOS-build RA machines. I can see it now:
mae-east.netaxs.com(config)# int f0/0
mae-east.netaxs.com(config)# ip icmp ra-response ?
ignore-just-at-the-local-exchange-point Just ignore from local ra servers.
ignore-from-all-known-ra-ips Ignore from any known ra server.
send-back-interesting-icmp-error-codes Try to confuse ra stats gathering.
dont-ignore-but-randomly-respond Ditto.
and:
mae-east.netaxs.com(config)# router bgp 4969
mae-east.netaxs.com(config)# neigh 192.41.177.166 withdraw-every-route-ten-times
mae-east.netaxs.com(config)# end
And they didn't explain well another point, which was that the millions of
route withdrawls per day (which we assume is a result of the Cisco BGP
algorithms) are not a problem which critically threatens the Internet as
we know it because (thanks to Sean and others) noone has noticed in practice
as increased CPU load on key routers due to dampening...
But anyway, Gilder does an enthusiastic job of refuting the Internet-is-going-
to-go-away-in-favor-of-usage-based-and-more-reliable-Intranets Metcalfe
hypothesis while simultaneously attempting to, um, arouse, the reader's
interest... To understand that last in-joke you'll have to read the article.
Avi
