Jim Strickland wrote:
Gang, the tcp-ip protocol was designed to survive a
nuclear war.
Packet-switching was invented by Paul Baran to provide a communications system
that could survive a nuclear war. IP is in principle is capable of being used
to implement such a system. However, there is nothing in IP that guarantees
that an IP-based internet will in fact be that robust.
The Internet as actually deployed is nowhere near sufficiently robust to
survive a nuclear war, or even a large-scale conventional war. While there
may not be single failure points that would be catastrophic, it would not take
the destruction of a very large percentage of backbone routers to take it
down. By "down", I don't mean a situation where no nodes can communicate
with other nodes. I mean a situation contrary to the normal experience in
which any node has a high likelyhood of being able to communicate with any
other node. The Internet isn't useful if you can only communicate with
an unpredictable subset of it.
Assuming
there is ANY connectivity going around a hypothetically disrupted/destroyed
major network hub, the packets should find it.
If all of the connections truly were routing peers, that would be true. But
many of them aren't. If a single major peering point (or a few of them) were
destroyed, the net would still function, and within days would be back to
business as usual. If 20% of the peering points were destroyed, the net
wouldn't be usable (in the normal sense) for quite a while, possibly weeks.