Diving into the Internet

Paying for the Net

by Joel Snyder

The problem is not bandwidth, rather how Internet access providers will charge and interconnect.

Networks are cheap. Information is expensive. Keep those facts distinct and you'll make it through the next few years without too much stomach acid. Don't get confused about how we're going to pay for the Internet of tomorrow.

Up on Capitol Hill, people who have made a business of owning lots of bits (such as Columbia, Paramount, and Ted Turner) want to have a cheap way to get the information to users so they can charge piles of money for it. Notice the distinction: cheap pipes, expensive information.

You don't want the expensive data? No problem. No one is going to make you pay to watch a rerun of "Gilligan's Island," listen to "Boring Hits Never Played at Woodstock," or read an article on miracle acne cures, all coming to you through your cheap 21st-century Infobahn connection.

Once someone figures out a good way to charge for information, we'll see the quality go up as the quantity explodes. Lots of folks are working on ways to integrate charging mechanisms with common Internet tools. So let's not talk about the question of how people will pay for the information they consume. It's going to answer itself. When there's lots of money to be made, inventive entrepreneurs have a way of finding the genies needed to answer those trivial technical questions.

What I want to talk about is the pipe; the wire. For the technoids among you, I'm talking about your SLIP or PPP or whatever-you'll-use-next-year connection to the Internet; the bare-bones hookup that jacks you into the network. The one that gives you data but no information.

How are we going to pay for the pipes? That's a big question right now, and it will only begin to be resolved by the time you read this.

The question of paying for the pipes came up because the great guardian of the backbone, the National Science Foundation(NSF), has discovered that it doesn't have to pour tax dollars into the network anymore to make it work. The Internet has so many users and has proved its usefulness so strongly that it no longer requires public funding.

The NFS is pulling out, mostly so it can give the money to someone else. In the wake of its withdrawal, the NSF has left an architecture for a completely different kind of backbone, operated by big telecommunications companies like AT& T and Sprint.

Naturally, this has everyone concerned; which is astonishing. We don't trust the government with our Clipper Chip decryption keys. We don't trust it to keep our handgun ownership permits safe. But we do trust it to set policy for the Internet; even more than we trust AT& T!

First someone suggested that we should all pay for e-mail services on a per-message basis. This plan was made up, I am sure, by someone who is used to buying data services from an X.25 network. These kinds of networks do charge you by the bit, and they're run by big companies such as AT& T and Sprint, so it seemed reasonable to assume that the same pricing model used in X.25 would be used for the Internet's new backbone.

Reasonable, that is, to someone who hasn't bought phone services in the last five years. Because; surprise!; telecommunications lines and equipment don't cost what they used to. Ten years ago I set up a network that was $1,400 a month for the connection and a nickel for every line of data someone sent or received. At the time, we thought it was a bargain. Heck, it was a bargain. But that was in 1984. Lines were expensive. Hardware was expensive. Even the software was expensive. No more. The equivalent service today is $100 a month for the connection and less than a hundredth of a cent for a line of data.

Networks are cheap. Today, the only sensible way to charge for network usage is based on the size of the pipe. You have a 9.6-Kbps SLIP line? That's one price, per month, all month. Faster lines cost more. Not only does this kind of charging closely mirror the actual costs incurred in running the network, it makes sense to the consumer of the service. That's a rarity in today's business world. In the Internet, reality meets economics and they both come to the same conclusion.

Consider the alternative: charging by the bit. First you'd have to completely redesign the protocols, re- engineer the hardware, and retrofit the culture of the Internet. When it comes to bits, networks waste far more than they use. Did you realize that when you telnet across the Internet, every single character you type causes 80 characters to pass over the wires? Even a $400 hammer looks cheap compared to that overhead. The "highly efficient" client/server protocols like Gopher are only somewhat better. A typical Gopher menu of 500 characters generates 14 packets (only 5 of which carry Gopher data) and another 1,500 characters of overhead. Which of those bits are you supposed to pay for?

The Internet is simply not designed to work on a per-bit basis. Frankly, the cost to perform accounting and billing on a per-bit basis through a network backbone is higher than the revenues you could collect for the traffic. The overhead cost in keeping track of a terabyte backbone is unthinkable. No one is going to stand for that kind of charging. If some bozo tried to charge by the e-mail message, bit, packet, or whatever, it would end up being so expensive that customers would simply take their bits and go elsewhere. Building large networks is not rocket science. We don't need some mega-carrier to run the backbone.

In fact, running a network is so simple that it has spawned a whole new cottage industry: Internet service providers. If you're paying some incredibly low fee for a SLIP or PPP account, you're probably buying it from one of the new small companies. These folks buy a 56Kbps or T1 circuit from one of the big providers such as Sprint, and then resell access to as many people as they can.

The business is a pretty good one. Invest less than $10,000, find 20 people who want dedicated connections or 200 who want to connect only occasionally, and you can pull in a nice $1,000 a month profit. You don't even have to quit your day job. Taco Bell franchisees, are you paying attention?

There's a problem with this. Bill Washburn, until recently the executive director of the Commercial Internet Exchange (CIX), called it "a policy-free vacuum." The problem is that under the new post-NSF backbone there are no rules. There are only assumptions. It is assumed that people in the Internet service- provider business will want to pass traffic to and from their customers. That's the way it always has been.

However, businesses don't like to work under assumptions. Businesses like contracts and agreements, and CIX, composed of businesses passing commercial traffic over the Internet, is not happy about what it sees. CIX members already have agreements with each other, but most of the cottage-industry, low-level providers aren't CIX members. CIX uses a "multilateral agreement" to keep the packets flowing. Join CIX, sign a single contract, and you're now party to an agreement with every other CIX member. Are we going to have to fall back to the days of bilateral agreements, where every single Internet provider is going to have to have a separate agreement with every other provider?

There obviously will be costs involved in operating the new backbone. How will they be divided? How will the small companies that are providing inexpensive access fit in? There must be agreements in place to ensure that every site on the Internet can communicate with every other site. Who is going to make those agreements? These are the key issues that will be decided in the coming months. It is possible to have a new backbone with equitable and reasonable costs for those who use the service. Building commercial and administrative structures to support the technology will require a lot of answers; -ones we don't yet have.

Joel Snyder (jms@opus1.com) is a senior partner at Opus One, a consulting firm specializing in telecommunications and information technology.

Copyright (c) 1994 by Mecklermedia Corporation. All rights reserved. Material may not be reproduced or distributed in any form without permission.