Related Web Works

[Star Tribune Online Business] Published Sunday, October 20, 1996 [Return to front] Is the Internet alive and well or headed [Return to Business section] for a slide?

Elizabeth Weise / Associated Press

The Internet is broken. The evidence is everywhere.

Outages drop millions offline for hours, sometimes days. The number of users has been doubling every year since 1988, and traffic on some long-distance routes doubles every four months. World Wide Web pages take forever to load because data pipes are clogged.

On the other hand, the Internet is doing just fine.

Audio and video applications formerly only dreamed of are now commonplace. Messages zip along 30 percent faster than two years ago. The major long-haul providers are increasing their capacity at a prodigious rate -- this year alone, MCI quadrupled the load its lines can carry.

Both of these outlooks are correct; the truth lies somewhere in between.

Engineers have been predicting the collapse of computer networks since before there even was an Internet, and the scare re-emerges every few years. It's a meme (Internet-speak for a self-replicating, self-perpetuating idea) that never dies.

The latest round had its start with a flurry of columns, speeches and interviews late last year by networking pioneer Bob Metcalfe, who warned of the Net's impending supernova.

"The Internet has outgrown its design and needs to be fixed," he said. "There are going to be more outages, and they're going to get worse."

Differing opinions

Metcalfe said he believes that the Internet as it is currently constructed, a loose, illogically connected bunch of computer networks, needs to step into the real world. But while he sees a system in the process of disintegration, others see innovation and growth.

"The Internet isn't going to fall apart. We're just in a state of transition to a faster, more stable network," says Fred Baker, president of the Internet Engineering Task Force, the body that sets Internet technical standards.

And as far as Internet demographer John Quarterman can tell, the system is working just fine.

"The trend is that service is improving, just not as fast as people are being added on. But that's the same as it ever was," said Quarterman, of Austin, Texas, whose Matrix Information Services provides some of the best figures for Internet growth and usage.

In a sense, the disagreement is between those who think the free market of ideas that makes up the global network of networks will work things out all by itself and those who say it needs direction.

Said Metcalfe: "I'm fighting the bio-anarchic intelligentsia that runs the Internet, the people who think it's alive and can't be managed." In a bombastic but bemused tone, he added, "I think we should hold a great big party, congratulate them for building such a wonderful thing as the Internet -- and then we should send them home."

To which Quarterman answers in the gentlemanly tones used by all involved in the debate, "Let me quote an old Texas adage: 'If it ain't broke, don't fix it.' "

Growing up

The network's growing pains are perhaps inevitable. Until last year, the infrastructure of the Net had a kind of de facto organization in that the National Science Foundation (NSF) ran the major backbone routes, the high-capacity long-distance lines that carried most network traffic.

When the government decided that the Internet -- originally, a Cold War research project -- was big enough to take care of itself, it relinquished its tenuous control and retired the NSF backbones.

That caused two things to happen: Suddenly, it was possible to run commercial traffic on the Internet (a no-no as long as it was created with public money), and a hodgepodge of commercial companies took over the long-haul business.

Commercial traffic meant more business and a land-rush in cyberspace as everyone scrambled to get online. It also meant millions more users every year and, consequently, an incredible demands for more capacity, or bandwidth, as it's known online.

Rather than one organization laying down fairly logical links between Internet service providers, a gaggle of commercial services put in the connections that worked best for their needs, but not necessarily the needs of the network as a whole.

To function reliably, a more centralized network is necessary, Metcalfe said, and he wants the companies that supply Internet connections to individuals and businesses to step in and take charge.

"The Internet services providers, of which there are 3,000 in North America, need to band together," he said.

In his vision, these free-market entities would meet regularly, track outages, debug the Internet protocols, rationalize the spaghetti of connection between networks and create a more regularized billing system between them.

In fact, he advocates using one of the Internet's own self-organized groups, the Internet Engineering Planning Group, to do it -- before other, less benign, entities decide to take it on.

"The current management vacuum creates the possibility for monopolies to step in and take over," he said. "One possibility would be the government, God forbid. Another, which is terribly worrisome, is the telephone companies. That would be death -- they're slow to innovate and they charge an arm and a leg. The third kind would be Microsoft -- and we don't want them running the Internet."

Actually, Microsoft agrees with Metcalfe. Peter Ford of its Internetworking group sees a future that holds some degree of consolidation, but in which different kinds and levels of service are available from many companies.

"The beauty of the Internet is that we'll always have options," Ford said. "It's like any other buying decision -- people can choose to buy their milk from Safeway or from a co-op."

Billing systems

That's good, because one of Metcalfe's calls for reform includes a move away from the flat-rate billing now common and toward billing for time and packets, the small particles of data every Internet transmission is broken into before being reconstituted at the receiving end.

"Flat-rate is a naive billing system because it means light users are subsidizing heavy users. If you're an Internet service provider, as soon as you get a customer, you have to keep him from using the service," he said.

But Quarterman noted that it was unlimited use that allowed for the experimentation that made the Internet the center of innovation it is.

"If we had had that kind of charging two years ago, we wouldn't have known what the World Wide Web was, because it would still be a researchers' tool," he said.

Quarterman said he believes more use will bring more customers who, yes, will use more bandwidth. But they'll also be willing to pay for it, which in turn creates economic incentive to build more bandwidth.

"It's not as if we put more cows on the Boston Commons and they ate all the grass," he said. "It's as if every time we put a new cow in, more grass grew."

Just as there is a lag between traffic increases and the building of new roads, commercial Internet services cannot add bandwidth exactly when it is needed, however. Metcalfe sees this as a dangerous situation that needs oversight. Others just see capitalism doing what it does best.

Another proposal anathema to many in the cooperative free-for-all that is the Net is to build virtual toll roads down the center of the backbone routes.

Currently, every piece of e-mail, every Web page, every audio file is broken down into thousands of packets, which wriggle through the network like spawning fish heading home.

One of the hallmarks of the Internet has been that, once in the pipeline, every packet had as much chance of getting where it was headed as any other. As a message bounced through five or 10 different networks as it swam home, it went as fast or slow as all the messages around it.

The basis for the system could be boiled down to: "You carry my traffic and I'll carry yours."

Unlike the telephone system, which gives users an "all circuits busy" signal when the lines fill up on Mother's Day, the Internet just slows down and drops packets. Its built-in correction system makes sure all the packets get re-sent and arrive in the end but, on a busy day, that can take awhile.

That isn't a problem for e-mail or even a slow-painting Web page. But for audio or video transmissions, slow service and dropped packets mean they cease to be useful. It is new applications like these that mean the old system is likely to change.

Under various tagging schemes currently under construction, all packets would be equal, but some would be more equal than others -- for an extra fee, they'd get to jump ahead. That in some ways would be a return to the ideas of old proprietary networks of the '70s and '80s, which died out when the Net came along precisely because of its egalitarian nature.

Bandwidth-hogging applications may bring them back.

"It's almost a religious issue online," said David Clark, a senior research scientist a MIT's Laboratory for Computer Science. "Many people are afraid that it will destroy the innovative edge of the Internet."

In the end, Metcalfe's gnashing of teeth is doing the Internet community a favor, Microsoft's Ford said.

"He's taking a very extreme position, but it's good," Ford said. "It's making people wake up and think and then hopefully act to make the network better."

Whatever the future holds -- brown-outs, minor collapses, differential levels of service, new payment plans -- the Internet seems uniquely able to evolve into just what's needed.

As one engineer said at the Internet Society's meeting in Montreal this year, "The Net is dead. Long live the Net!"

[Return to Business section] © Copyright 1996 Associated Press. All [Return to front] rights reserved.