Internet Myths and Public Policy

This essay is based on comments I wrote for the Email for All discussion hosted by the Markle foundation. You can look at http://www.iaginteractive.com/emfa for background on that discussion. There was much confusion in the discussion between the Internet and certain applications, such as Email, which run on top of the Internet. It is important to understand this distinction in order to formulate effective policies.

The Importance of Understanding the Internet

Myths vs. Realities

While I am enthusiastic about email, using it for over 30 years has given me some perspective.

To be very direct – "Email for all" misses the point. It is a product of an old style of thinking that focuses on the application rather than on the enabling technology. The Internet (and the PC) have fundamentally changed the rules of the game by providing technologies that are not limited to preconceived notions of their purpose. Understanding how and why is a prerequisite for any informed discussion on email policy.

Yes, you will get universal connectivity. It's no big deal, just one of the many capabilities that emerge from IP infrastructure. But you will only get it if you learn to take advantage of what is available. Email, for example, needs a PC or similar device for full capability yet messages can be adequately delivered with little new hardware, even by a standard telephone.

Despite having been involved in the field since the mid 60s , I'm more excited than ever about the implications and the possibilities of email. Rather than being futuristic fantasies, the discussion of email is now about actual implementations and applications.

To fully appreciate the implications of trying to provide "Email for all" , it's necessary to have a rudimentary understanding of what the Internet Protocol is. And what it isn't.

We can start by going back to the key decisions from the 1970s which created an IP (Internet Protocol) infrastructure that could evolve rapidly and independently of the particular applications that use it.

The existence of a simple protocol designed to interconnect networks created an environment in which creating the web was a matter of establishing conventions for naming items on other systems in terms of the system's Internet name and the local file system name of the item. Add a markup language (HTML) and we have the web. If that doesn't seem simple enough, remember that one can write HTML documents using a simple text editor.

At a technical level, an IP message consists of a destination address and data. Drop it on the network and it quickly appears at the destination elsewhere in the world. And sometimes it doesn't since the network is not perfectly reliable.

The power lies in the Internet's indifference to what it is transporting, "bits is bits". It doesn't presume to even understand what reliability means for a particular application. This extends to not imposing the cost of guaranteed delivery on applications such as streaming audio timeliness is more important and perfect delivery.. What seems at first to be a source of unreliability is really the strength of the service.

This is counterintuitive. To put it another way. By avoiding the false promises of guaranteed delivery at the transport layer,,the applications are forced to take responsibility for their service. Packet loss is a trivial issue compared with programming errors, maliciousness and the many other issues that an application must handle. By taking a broad view of failures (packet loss, programming bugs, maliciousness etc) we are able to create global services without central control or coordination.

As an aside, TCP (Transmission Control Protocol) is simply a protocol used by applications to get a reliable connection if they want it, and most do, thus the shorthand of TCP/IP, but I am careful to say just "IP".

Given all these, we need to reexamine some "obvious truths":

Myth	One must make efficient use of resources.
Reality	Efficiency is itself a cost. By making the network itself very inexpensive, we can focus our efforts on creating new capabilities and services.
Myth	The applications dictate the network requirements.
Reality	The generic network not only enables existing applications, it allows for the discovery of others.
Myth	The network is a vital service that must be reliable and predictable.
Reality	As noted above, such reliability is the responsibility of the applications, not the network itself. The network cannot make up for unreliable applications but applications can provide reliable delivery on an unreliable network.
Myth	We must provide subsidies and funding so that everyone can afford to be connected.
Reality	This is a dangerous myth, since it has the effect of assuring that the services are expensive in that it removes the incentive for competitive pricing.
Myth	Network wiring is complex and expensive.
Reality	Now that there is interest in consumer networking, the marketplace (i.e., individuals like myself) is working to make it simple by including, for example, the ability to use existing wires. Windows 98 already does automatic address assignment.
Myth	The Internet is dangerous – we must provide firewalls and filters to keep our computers safe.
Reality	Reality: Old systems and LANs have policies that assume a very benign environment. But making it safe to connect them to the larger Internet is no different that making it safe to connect a phone to the phone network without having a plug-board PBX in the way. To oversimplify, one problem is that current personal computers leave all the "doors" (access) open by default and thus let anyone get at all the data in these systems.

Public policy discussion of the Internet is really about the applications that take advantage of the infrastructure. The network infrastructure just caries bits and has no information about what the applications are doing. Policies can be aimed at how the Internet is used but it is not meaningful to try impose such social policies on the infrastructure itself. At least, not without cause frustration and disrupting the economic benefit.

Implications

We can see the effect of regulatory policies in the current telephone network. In an attempt to provide Universal Service, we have created a structure that presumes telephone calls are expensive, and then ensures that they stay expensive by taxing connections and using the money to pay the phone companies.

We have a situation in which a copper wire can carry megabits of data and thus dozens if not hundreds of phone calls. But there is no incentive to take advantage of this capacity. Why is this situation tolerated? Because in pursing the goal of making phones universally available and affordable, we have done precisely the opposite. We have ensured that they are expensive and that new services can only be introduced if the providers deign to implement them.

Despite the attempt to legislate it, Internet Telephony in itself is just a way of using the IP infrastructure. There is no requirement to use any particular approach. The current "standard" is just a convention one can choose to use.

Yet, the Universal Service fund has become accepted to the extent that there are serious proposals to tax all connections between the traditional phone network and the Internet on the off-chance that someone may use it create an audio path to someone else via the Internet. If this seems strange, it is only because it is.

All along, the intentions are the most honorable. And it gets worse when we add another noble goal to the mix – connecting schools to the Internet.

It is indeed important that we take advantage of the Internet, to advance education. But connecting to the Internet is a small part of the larger problem of understanding information and computation. Are we teaching about databases; about information representation; how word processors are more than just typewriters?

But those are difficult issues. It is easier, and still useful, to focus on the physical infrastructure. Getting a connection to the Internet is important. But the schools that can't fund such a connection have more serious problems, such as not having enough electrical power to the classrooms. In order to address this, we need to face up to funding it.

If we can't get the local governments to recognize the value of education, it is tempting to provide federal funding without an explicit "tax". Universal Service fund is too tempting a pot to ignore.

Universal Service?

So, we've created a fund based on the assumption that telephony is still very expensive, and we use it to ensure that high prices for phone calls will persist. And now we want to do the same to schools . While the marketplace is rapidly working on novel approaches, such as using existing wiring and wireless connectivity to essentially eliminate the cost of Internet connections, we want to do to the Internet what we have done to telephony. Some of the money will likely go into needed structural repairs, but the damage done by creating an hourly cost on Internet connections will do far more damage.

A meter ticking while online will ensure that we do not have universal access to the Internet. Well, not to everyone, just to the poor who are most vulnerable. And, while access from the schools is important, those without Internet access at home or nearby will continue to be disenfranchised.

The fact that local telephone calls have generally been unmetered has been worth trillions to the economy. Like the Interstate Defense Highway System, it provides the basis for connectivity. In the United States, unlike the rest of the world in the grips of the PTTs (governmental monopolies on Postal, Telephone and Telegraph), we've created a thriving online industry which has become the Internet industry. We were already online with modems ready to rapidly move from proprietary services to general Internet connections.

And now, in the interests of assuring that no one discovers that "long distance" telephone calls don't have to be expensive, we are willing to sacrifice a major driver of our economy. For those who are excited about using the technology to create new communities and new services, there is a punitive tax.

This isn't to say that all is well with the Internet. It got popular before it was finished. We must transition from the IP version 4 to version 6 so that we don't have the artificial scarcity of addresses that makes setting up networks more awkward than necessary. And we must address the technical issues of access control that make people and companies reluctant to directly connect PCs to the Internet. Firewalls, proxies and other such mechanism are not solutions. They are temporary work-arounds that make the Internet much more difficult to manage and introduce some of the administrative control points that frustrate effective use and effective competition. Eliminating impediments to the simple IP layer is important.

Making the Internet itself a relatively pure marketplace doesn't require that one be a Libertarian. Social policy can be implemented in a positive way by creating services and capabilities that take advantage of this infrastructure. But the IP layer itself has thrived by being a commodity marketplace.

Summary

The main points I want to establish are:

Don't confuse the IP infrastructure with the applications that use it.
Email is just one of the applications that benefit from the IP Infrastructure, and email itself is just part of the larger spate of messaging services that are emerging.
The Universal Service fund and its cousin, Universal Access to the Internet, is are likely to cause damage. We do want everyone to get the benefit of the Internet, but doing so by assuring that connectivity is expensive seems self-defeating.

For those who want to learn more on the end-to-end principles, you can read the classic paper on the subject: http://www.reed.com/Papers/EndtoEnd.html. There is also a more recent paper which focuses on how this affects ATT's business http://www.isen.com/stupid.html. There is also my more recent essay on the IP Infrastructure.