Metcalfe's view Predicting the Internet's catastrophic collapse and ghost sites galore in 1996 - Metcalfe's original end-of-Net column. Bradner is a consultant with Harvard University's Office of Information Technology. He can be reached via the Internet at sob@harvard.edu

By Scott Bradner
Network World, 11/18/96

The sky is not about to fall. One of the most common reactions in the press and, to some degree, the technical community, to Bob Metcalfe's predictions of the impending collapse of the Internet, has been to treat him as a Chicken Little - running around in circles loudly lamenting the end of civilization as we know it.

It is a nice, simple concept, easy for the press and public to understand - the Internet will be toast by Tuesday (or was that Thursday). But there are two things wrong with the story. First, it is effectively impossible for the global Internet to fail in a catastrophic way. Second, Bob talks about problems that might affect large numbers of users - not quite the same thing as a systemic collapse.

The Internet consists of hundreds of thousands of interconnected networks, each under its own management. Most of these networks are within end-user organizations and thousands of small Internet service providers. In addition, there are a few dozen large ISPs, each with a service area spanning geographic regions and, in some cases, national borders. This structure is a more richly endowed version of the global telephone network structure (Note that with the global trend towards telecommunications deregulation, the structure of the global telephone network is getting to look more like the fragmented Internet).

Is a systemic collapse of such a system possible? No, but some pundits try to claim that it is. They come up with nightmare scenarios where all the ISPs get routing updates that their routers see as some form of death pill, or all ISP routers are simultaneously sabotaged by some evil force. Some pundits even point to the meltdown of the telephone switch signaling protocols a few years back as evidence that the same kind of thing could happen to the Internet.

Is systemic collapse probable? No. The Internet is far too diverse these days.

I do have to agree with Bob that there will continue be vendor- and facility- specific outages. As some of the vendors continue to expand their customer bases, the effect of outages will grow (megalapses, if you will). But how do these outages compare with the outages experienced by other service providers? A few months ago, a major failure of the power grid left most of the western U.S. trying to communicate by candlelight. And on an average day, more than 30,000 people in the U.S. are without telephone service for an average of five hours each (See http://www.atis.org/atis/nrsc/whatsne2.htm). Yet no one points to an impending failure of the power or telephone systems.

There is a problem in that many of the existing ISPs are undercapitalized or do not have adequately trained staff. You can expect such vendors to experience episodic overloads, resulting in slow or unreliable service. But, just as I changed my telephone long-distance provider when I had trouble getting calls through because of overloaded lines, I expect users will change ISPs if they are getting poor quality service.

Bob has been engaging in more than a little bit of hyperbole to try to get the attention of the ISPs, but what I find far more troubling than Bob's message is the way it has been simplified and magnified by many in the press. It's not that the press is trying to create controversy. Rather, I fear, it's because too few in the press understand the technology they are reporting on. If it all seems chaotic and magical, then it is easy to think that the magic may end.

Disclaimer: Harvard has been on the Internet since before we knew it was the Internet, and the university has seen quite a range of service qualities over the years, but the above are my own opinions.