the Internet can not exist

By: Scott Bradner

It has been said that according to aerodynamic theory bumblebees can not fly. In the same sense, according to all kinds of theories the Internet can not succeed, even to the extent that it already has and if we look at the economics of the Internet, it should not exist today. In other words, the Internet must be a figment of our imagination.

The Internet is quite different than "traditional" data networks. The most traditional of data network technologies is IBM's Systems Network Architecture (SNA), that technology as well as Digital Equipment Company's DECnet and most of the data services that telephone companies have offered in the past (such as frame relay) and expected to offer in the future (such as ATM) are designed to provide specific guarantees of network performance. The TCP/IP protocol on which the Internet is built was not designed to provide the same kind of guarantees. There is quite a bit of work currently being done with TCP/IP to add the ability to provide some degree of service guarantees (RSVP for example) but there is no technology currently in place in the Internet that can provide service guarantees.

This is a real problem to many people with backgrounds in traditional networking. Some SNA consultants have gone so far as to say that you "can not build" a corporate data network with TCP/IP because business data networking requires reliable service guarantees. Yet more and more companies are migrating the applications that they developed on SNA and DECnet networks onto the Internet and TCP/IP. These companies have made the decision that the flexibility of the TCP/IP protocols outweigh the uncertainties in delivered service levels. A decision that the pundits said would never happen.

A second reason that the Internet should not exist is because it was going to be replaced. A few years ago there was a world-wide assumption that the next generation of data networking was going to be the Global Information Infrastructure (GII). This GII was going to be omnipresent and provide wonderful things but it was going to replace the Internet since the Internet was this toy that the scientists played with, not something that was well enough planned to be the backbone of world communications. Funny thing, the Internet is now living what the GII could only dream about.

Another problem is content, according to expectations there is not much on the Internet. One of the features of the GII was that it brought information to its users. But the information was going to be from a few large content providers. The information could be the contents of today's newspaper but in most of the descriptions the information was actually entertainment. Movies played on user demand for example. Almost none of that happened - according to that picture there is little content on the Internet. What the GII people did not understand was that the Internet showed that there would be millions of small information sites (currently represented as web pages) not just a few big ones.

The GII supporters also did not seem to understand that I might want to communicate to my friends over the network. Email between GII users was rarely mentioned as a reason for building the GII and yet email is one of the biggest uses of the Internet - well behind the web in popularity but far ahead of video on demand.

Another reason that the Internet should not exist is scale. According to the pundits, TCP/IP would never be able to support millions of computers on a network. It would run out of addresses or collapse because it was designed at a time when a very big network had only a few hundred computers attached to it. Well, the IETF came up with a way to extend the address space of TCP/IP to support billions of hosts and it turned out that IP works quite well with millions of computers on the same net.

Also it was felt that TCP/IP could never run at high speed. A major executive in IBM once said that TCP/IP would never run faster than a few million bits per second. This has not quite proven to be the case - some current Internet backbones run at more than a billion bits per second and TCP/IP communication has been demonstrated at hundreds of megabits per second between two hosts. TCP/IP now runs surprisingly well on links as slow as 2400 bits per second and as fast as 2.4 billion bits per second.

But the biggest reasons that the Internet should not exist are organizational not technical.

The pundits have said that businesses would not use the Internet because there was no on in charge. There is no place they could call to get things "fixed" when they broke. But this worry has always struck me as very strange. The implication is that with other services that business deal with there is "someone to call." But who do look to fix things if your phone call fails to get through? Most of the time the problem is "somewhere else", not in the control of your local telephone provider. There is no global telephone control point, why should anyone expect that there would be one with data networking. In fact, the Internet service providers are consolidating faster than the old monopoly telephone companies are breaking up and getting competition.

Finally, the pundits say the Internet would fail because it was brought to you by amateurs and not by the professionals in the telephone companies. Well, as the Internet has developed the amateurs seem to have been doing a better job than most of the telephone professionals. But the phone professionals are learning and may catch up soon.

So the Internet is in your Imagination. It can not exist because it violates all of the expectations that the professional prognosticators had for the future. I, for one, like this imagined world.

Scott Bradner