Here’s where you’d ordinarily see our weekly roundup post. Unfortunately, I haven’t been able to get a web connection to my provider for the past several hours (though they’re definitely up and running); all attempts seem to stall while trying to pass through a downed machine in Chicago. Isn’t the whole point of the Internet supposed to be that there are multiple paths from everywhere to everywhere so this kind of thing can’t happen?
Be that as it may, I am logged into a shell account and posting this via lynx (if you don’t know what that means, you’re probably not as old as me), and the interface is far too painful to type anything substantive. So I’ll plan to post the usual roundup sometime tomorrow, after they’ve cleared the gunk out of the Intertubes.
Should we expect a post on fractals this week? Unfortunately, Benoit Mandelbrot has died: http://www.nytimes.com/2010/10/17/us/17mandelbrot.html?_r=2&src=tptw
When the folks at the Advanced Research Projects Agency (ARPA) were designing internet technology for the military, it was intended to survive attack by an enemy with atomic bombs. The design goal was probably something like that it would take more than N destroyed nodes to partition the network into smaller nets, with some adjustments for cost of losing certain subnets and probability of attack on each node. It was a military-grade idea.
That network may still be out there, but the Internet isn’t it. A multi-level hub-and-spoke design, with big hubs connecting smaller hubs, and so on ad infinitum, is more cost efficient. Of course, if one of the high level hubs goes down, everybody downrange loses their connection.
The top level routing hubs, several hundred of them, do have a fairly high degree of redundancy. In part, this is because there are several competing network companies connecting them, and while the hubs exist primarily for peering–passing traffic between customers on two different networks–they also have agreements to carry each other’s traffic in emergencies, overloads, etc.
Web sites that are very sensitive to downtime can have their servers located at data centers at the peering point hubs. One of my clients uses this kind of hosting and in eight years I’ve never not been able to reach them because of a problem on their end. For this level of service, they pay around $1500/server/month. Onthe other hand, your website’s host advertises a shared plan for $6.95/month.
BTW, back at the height of the browser wars, when the browser makers were encouraging web masters to put “Optimized for Internet Explorer” and “Optimized for Netscape Navigator” badges on their sites, my personal site was one of the few, the proud, the “Optimized for Lynx.”
It’s actually a myth that the ARPAnet was designed to survive nuclear attack. It comes from some offhand, after-the-fact quotes. ARPA’s goal wasn’t to design technology for military use per se; rather, it was a branch of the defense department whose goal was to fund “advanced research” into anything interesting, in an open-ended way. The idea was that instead of the USSR surprising us with their technical innovation, we’d surprise them.
Well, I’m pretty sure the military was interested in ways to maintain communications during and after a nuclear war, and we know ARPA was researching networking technology that could survive the loss of nodes or links, but now you’re saying these two things were unrelated efforts? Let me tell you, I’ve done some defense contract work and that…sounds very likely, actually.
Maybe I should say that the TCP/IP internet could be set up for high survivability, but isn’t.