The Internet After the Fad

Remarks of Dr. Robert Metcalfe at the University of Virginia
May 30, 1996

Dr. Robert M. Metcalfe began his publishing career in March 1992 by joining International Data Group as publisher and chief executive officer of Infoworld Publishing Company. He has served as IDG's vice president of technology since 1993.

Just before joining IDG, Dr. Metcalfe was a visiting fellow at the University of Cambridge, England, and was writing for Computerworld, Communications Week, Digital Media, Network Computing, and Technology Review. He now writes a weekly column for Infoworld, where he has served as executive correspondent since 1994.

Dr. Metcalfe is a graduate of the Massachusetts Institute of Technology and received his Ph.D. in computer science from Harvard in 1973. He taught part-time at Stanford for eight years, ending in 1983 as consulting associate professor of electrical engineering. He is a member of MIT's Board of Trustees.

In 1972, Dr. Metcalfe went to the Xerox Palo Alto Research Center, where he invented Ethernet, the Local-Area Networking technology that now connects almost 20 million computers worldwide.

In 1979, he founded 3COM Corporation, a computer networking company which over the next 11 years he helped to grow to $400 million in sales. At 3COM he held various positions including chairman, chief executive officer, president, division general manager, and, most memorably, vice president of sales and marketing.

In 1988, he was awarded IEEE's Alexander Graham Bell Medal for his work on the invention, standardization, and commercialization of local area networks.

Dr. Metcalfe is married to author Robyn Shotwell and they have two children.

Table of Contents

The Collapse of the Internet

Characteristics of a Collapse

Potential Causes of Internet Collapse

Service Providers Responsibility for Continuing Service

Lack of Relation between Supply and Demand

The Growth of Intranets

Metcalfe's Law

Partitioning the Internet

The Cost of Bandwidth

Is Metcalfe a "Chicken Little?"

The "Available Excess Capacity" Argument

Expansion of Computer Capability compared to Expansion in Bandwidth

Significance of Deregulation

Forms of Competition

Two-way Television

Historical Parallels

Data Compression

Transition to Metered Usage

Need for Quality of Service Guarantees

Need for Encryption

Need for Better Audio and Video Capability

Need for Zoning Restrictions

Expanding Internet Access

Improved Educational Use of the Internet

Instituting "Micro-Money" Transactions on the Internet


Collaborative Filtering on the Internet

The Role of Editors and Publishers on the Internet

Emergency of more Community Content in the Internet

Institutions hurt by the Internet

The Role of Government

The Long-Term Impacts of the Internet

Solving Internet Problems

The Internet is overdue for a major collapse. This might not necessarily be a total failure. It might take several forms. One is that problems might lead to a collapse in confidence. Companies would then take a lot of their information, which they have been putting out on the Internet, and sort of cautiously easing out into their Web pages -- might pull it back behind the firewall in an overreaction to some security breach. But if everyone sits behind firewalls, then there is no value anymore. In other words, the value of the Internet is related to the information which is in front of the firewall, out where everyone can get it.

Another kind of collapse would be an advertising collapse. Advertising revenues should be and are becoming an increasing fraction of the monies which finance the Internet. In the advertising business, you measure the media that you buy ads in. And as a medium, the Internet is not well measured. And so, the traditional channels of advertising purchase are buying ads on the Internet now experimentally.

The tools for measurement are being put in, but those tools may have an unpleasant result. They may discover who is actually on the Internet and how little they buy and what a bad advertising investment they are.

In which case, those advertising dollars, instead of growing, as they are right now, might begin to collapse. Another form of collapse.

And just before getting on to my central theory of collapse, there is the growing excitement, press and so on about audio, video and Java, which promise to dramatically increase the amount of traffic on the Internet at a time when people are complaining about the Internet. Which causes me to ask the following question of you:

How many of you are currently dissatisfied with the reliability and/or performance of the Internet? Would you raise your hands?

Well, I want you to know you are not alone. I am very confident now when I ask that question. There is usually one person who refuses to raise their hand, and they work at Sun Microsystems, and they have a T-3 channel right to their desk. But even they are a little -- because they have come to realize that having a T-3 channel to their desk does not solve the problem out there, although it is a lot better than 14.4.

So the coming collapse of '96 and collapses of '96 -- and as I said earlier, they may already have occurred -- the Internet has the slows -- there is no denying that. And these slows have been termed "brownouts." The term "brownout" comes from another term which preceded it, I believe. They were called blackouts. And then a brownout was sort of a mild blackout.

Well, on the Internet, it is happening the other way around. We are getting the brownouts, and the blackouts are ahead, like the blackout of -- was it '66, in New England? Because this is Virginia, and you did not have the problem, but we had the problem where the lights went out.

That is the kind of collapse I am talking about on the Internet -- not a brownout, but a collapse -- meaning an outage that lasts a long time for a lot of people. And here is where we get into semantics, when the defenders of the Internet will not be satisfied. If a million people were out for an hour, they would not call that a collapse. I would call that a collapse.

Characteristics of a Collapse

If 100,000 people were out for two days, would that be a collapse? What constitutes collapse? It will be a matter of degree.

The FCC requires telephone companies to report outages of 50,000 people or more for an hour or more. That is an operational definition of impact.

I mean something really bad, not 50,000 for an hour. Much worse than that.

I have asked the readers of Infoworld to send me reports of collapses, without defining the word "collapse." So I am trying to learn what our readers think a collapse is. And they do not -- what I am learning is they do not make this distinction that I am making between a brownout and a blackout. For many of them, the brownout is collapse.

But that is not what I am talking about. I am talking about a more catastrophic event.

So we are having these outages. There is no place where the outages are kept. You know, like when you run a computer system, there is a place where outages are reported, and you go and you can go see them. And there are repositories of outages spread out around the United States, but there is no place you can go to find -- it is because the Internet is so resilient and biological and distributed and anarchic that there is nobody -- there is no hardball IS organization trying to track it and debug it and TQM it. Which is one of my complaints.

Potential Causes of Internet Collapse

There are four potential causes of collapse. First is the capacity problem, over time, leads to gridlock. Already now there are peak periods on the Internet, where the packet loss rates through the backbone exceeds 10 percent. Ten percent is a huge number.

In the early days of Ethernet, occasionally an Ethernet -- it is very rare -- in fact, I do not know of any time that it has ever happened -- but an Ethernet will lose a packet. And 1 in a 1000 -- I mean I had people lining up to lynch me because 1 in a 1000 would get lost on this Ethernet because of generally -- who knows why? No one knew why -- it is such a low rate.

Right now, during peak hours, you can find statistics in the routing arbiter database that show packet loss rates exceeding 10 percent during busy hours. So the first kind of collapse would be where packet loss rates went up substantially due to lack of capacity in backbone. And capacity is not a scalar, it is a vector.

It is -- capacity is a spatial phenomenon. And demand and supply of the Internet are distributed in space and time. So these overloads occur. They move around. In fact, there is a gentleman named John Quarterman who actually has developed videos that show the -- he calls them storms -- delay storms moving through the Internet.

I should tell you, Quarterman disagrees with me -- that there is a collapse coming. But his data suggests that there is a collapse coming.

Another cause of collapse would be equipment outages, like the backhoe problem. And last week, MCI lost -- I am not into the details quite yet, but they lost an FTDI ring and put out a good portion of the West Coast for an hour and a half. That would be an equipment outage. I do not think that was a backhoe, but something broke.

The third thing would be software bugs, which is that the bugs in the routers in the back of the network have bugs in them. I do not know what the bugs are. I just know that there are bugs in them. How do I know that?

And a lot of bugs, as you know, are loading dependent. That is, in the lab, the tests were not all done in every possible way, but under heavy load, often the bugs are excited. And there are reports of these bugs already. There are some well known bugs, like the router flaps, where an oscillating circuit -- the response of the distributed routing algorithm to a circuit going up and down -- causes the traffic in the network to slosh around like the water in a big tub.

Finally, there is sabotage, the fourth possible cause of backbone collapse.

So here is what I think we should fix. We should fix the poor management. The Internet was not designed for its current uses. And even the regimes under which it is managed are outdated. There is insufficient measurement being done. There is no central reporting of outages. There is little commitment. In fact, there is ideological hostility to sound management practices of the Internet backbone.

I usually get attacked at this moment.

Service Providers Responsibility for Continuing Service

I have begun reading IS -- there are about 1,500 or maybe 2,000 Internet service providers in the United States. I have begun reading the contracts that they ask you to sign when you buy Internet service from them. These contracts are hysterical. All of them are a denial of responsibility on the part of the Internet service provider.

There are exceptions. One Internet service provider is willing to offer you a refund if your outage lasts more than 24 hours. I saw one of those.

And I saw another one -- UUNET now has a service where they meter your use of the T-1 circuit and, on a moving average basis, sort of bill you for what you actually use of it. Now, this is an ideological departure from the Internet, where the bio-anarchists who run the Internet believe that you should only charge for capacity, not actual use. And this is ideology. It is a bad idea.

Anyway. So the ISP contracts. One of the ways we can fix this problem is get the ISP contracts to offer guarantees or at least recourse in the event that bandwidth is not delivered or packets get lost or delays get too long or -- in particular, you might want to ask your ISP what facilities your ISP is buying from his ISP's to support your usage. How many T-3 circuits is he buying to support it?

It's laughable when you buy a T-1 circuit and he has a 56 KB circuit going off upstream. So maybe you might, in your contract, want to say that the ratio of upstream circuits to the customer base should be such and such.

Lack of Relation between Supply and Demand

The fundamental problem, I think, is that there is no effective communication channel between supply and demand. That is, it was not built for this purpose -- the Internet was not. And the revisions to it -- the whole ideology of the Internet is built -- is designed so that it is very hard to make this argument in an Internet forum -- the notion of metering. That metering is one of the elements in a communication channel between supply and demand.

As demand goes up, that should cause the Internet to grow, so we will not have a bogging down. But the bogging down is caused because we do not have a communication channel between demand, which is going like this, and supply, which is who knows what.

QUESTION: Who is responsible for keeping the Internet from collapsing, the Federal Communications Commission?

MR. METCALFE: Well, not the Federal Communications Commission. I thought I found the body, actually. I made an error. But let me describe the error. I went to the Internet Society, who I thought would be a good choice to organize the management of the Internet. And in their Web pages is a thing called the Internet Engineering Planning Group, which is a sister organization to the Internet Engineering Task Force.

And we do trust the Task Force to design the standards and everything for the Internet, as quirky and organization as that is. Why not trust the Internet Engineering Planning Group to do it?

So then I went down into their pages, and I found that the latest postings or files under there were March of 1994. So, hmmm, what does that mean? I did not know what it meant.

But then under IEPG, there was a group called NANOG, the North American Network Operating Group. And that is my answer to your question. IEPG or NANOG or -- NANOG is active -- for small values of active. But that would be the appropriate place. That is why IEPG would be preferable, but it -- the activity seems to be at the NANOG level. But you are right, I think it should be international, aimed -- but, anyway, I focused on NANOG, at least just worrying about North America.

And, you know, the Web is supposed to be wonderful. So I rely -- I foolishly relied on the Web to begin to learn about NANOG. And having found it the way I did, I happened to write -- the Internet Society was a good place to get this management process started. And it had an organization called NANOG to do it. And then I started complaining about NANOG, like it does not meet regularly enough, it seems to be dominated by -- at least on the on-line forum -- it seems to be dominated by cranks and flames and all that stuff.

The rebuttal I got in a torrent of hateful E-mail was not a rebuttal to any of my complaints, it was that I had made the mistake of thinking that NANOG had anything to do with the Internet Society. We're completely independent. Which is, I found, very reassuring: NANOG is not even related to the Internet Society.


But that is the kind of answer I think we need. And it is cooperative. The FCC should not be involved. The FCC is going to get involved, but we should fight that. But something like NANOG, only being taken more seriously and with a greater level of investment.

The Growth of Intranets

Now, the response that the community of Internet users seems to be making -- the growing community -- I think is the wrong one. And that response is the development of things call intranets. Now, there are two intranets. You have heard the term "intranets"? In short, that is the application of Internet technology to corporate data systems. These are the building of separate internets related to the Internet -- through firewalls, typically.

There are two reasons I think people are building intranets. The good reason is that Internet technology is turning out to be great for solving internal information systems problems of corporations. That is a good reason. The other reason people are building intranets is that the people who run these corporate information systems are conservative, because they have responsibilities. And they do not want to trust their company's information to this insecure, unreliable, declining performing thing called the Internet. So they are building their own internets.

Now, this is not, long-term, what we want to have happen. And lest you think it is just the crass elements of the private sector that are doing this, I have just heard within the hour that we should built another Internet for us, for the universities, so that we can get the performance we deserve, instead of that junk out there which is caused by business and the consumer stealing the Internet from us, which we built and we are entitled to it, and they are taking it away from us, so we are going to build our own.

Metcalfe's Law

These are all incorrect responses, because of a law. I made this law up. It is called Metcalfe's Law. I did not call it Metcalfe's Law, but I am delighted that other people are. And it says that the value of a network grows as the square of the number of users. Well, if that law is even remotely true, subdividing the Internet, for whatever reason, cuts into its overall value. Because it inevitably detracts from the connectivity that we are seeking.

So rather than build our own separate intranets, we should be fixing the Internet. That is the general premise.

There is another term I would like to introduce called extranets. Latin scholars will recognize "inter," "intra" -- it struck me that we needed to come up with some meaning of the word "extranet." Plus, if you say Internet and intranet a lot, and if you have the slightest accent, no one can tell the difference. So extranet means the application of Internet technology, say, in a company for reaching outside the institution -- extraterrestrial, extracorporate, extra.

And this, I think, is the growth area for business on the Internet -- not just using Internet technology for internal business systems, accessing employees and internal systems, but using the Internet for getting out to customers. And the consumer is often the customer. So I think the potential in the Internet, and especially for business, is in the extranets, not the intranets. The intranet is just a temporary, low-hanging fruit in reaction to the poor performance of the current Internet.

Partitioning the Internet

QUESTION: Wouldn't partitioning the Net solve the problem?

MR. METCALFE: I would disagree. Pooling is preferable to partitioning. It is not a big thing, but it is an incrementally bad thing to partition resources. You lose the economies of scale, the benefits of redundancy, economies.

And what I think is happening -- well, another reason, and which you may find more persuasive -- is after the universities build this other Internet, then it will begin to degrade in exactly the same way, because it has the same DNA as the Internet which is currently declining. So you will delay the day and waste a lot of money. And then it will be bad in the end, too.

But I think it is more fruitful to go find out what is wrong with the Internet, why it cannot respond to increasing demand, and fix that, and then pool resources instead of partitioning. That would be the argument.

The Cost of Bandwidth

Now, one of the -- and I apologize to any of you who are bio-anarchists and disagree with me about this -- it just a term I invented to really upset you so we can have an argument -- but the bio-anarchists argue that bandwidth is free. So, for example, any effort to meter or charge for it is old world, pre-paradigm thinking. And if you really want to insult someone, just tell them that they just don't get it. You know, these young people walk around and they say, You just don't get it, do you? I guess not.

Anyway, bandwidth is free in the new paradigm. Like memory is free. You know, like memory used to cost a dollar -- we just heard from Jay Forrester yesterday. He paid a dollar a bit per month for the memory in his computer in 1957. And now, memory is extraordinary -- is free by comparison, except memory still constitutes a major cost -- a major fraction in the computer systems we buy. Why is that?

Because our appetite for memory has expanded more rapidly even than the economics of memory -- or at least kept pace. "Grove giveth and Gates taketh away" is the expression.


I do not think bandwidth is free. It is not going to be free. And it is going to be scarce forever. And, therefore, we should treat it as scarce. Even though it will be freer than it is today, it will always be a factor and we should therefore treat it as an economic good and create a market in it. And, guess what? We should meter its usage.

And the trend is inexorably toward that. We should just hurry up. Because the lack of metering is one of the effects that is leading to this bogging down. Because there is no -- you know, once you rent out this 56 K line to this customer as an ISP, it is then in your economic interest to install as little as possible behind that line to support it. And, really, it is in your business interest for that user not to use the line that you have sold them -- which is really odd. You would like them to.

You know, in a more rational business environment, you would expect that the more they use the line, the better that would be. But once you have sold the capacity for a fixed price, all the incentives are wrong. So metering. And then a special form of metering, which is settlements. You buy Internet service from an Internet service provider who, in turn, buys it from an Internet service provider, who, in turn, in turn, in turn, all the way on up to the NAPS, the network access, the big ISP's in the sky.

And right now, there are no settlements agreements among those ISP's. They just connect the lines between them on some sort of very crude agreement: "Aren't we engaged in a paradigm-shifting activity and don't we feel good about this, that we are hooking ourselves together?" There is no business agreement there that says if I dump 17 terabytes of data into your port, like, I should pay you for it. There is no settlements agreement.


And there should be settlements. And that is another form of metering.

Now, I am almost certain now that this is -- these two positions, which are rejected by the bio-anarchists -- I am almost sure this is a correct argument, because of the response I get. I am learning how to detect disagreement and weakness in arguments.

Is Metcalfe a "Chicken Little?"

I do not get arguments back that say, The capacity of the Internet is 18 terabyte per seconds per square meter, and the current demand is only 12.3 terabytes per square meter. Hence -- and the projected growth rate is this. And therefore there is adequate capacity for the next 12 years. No one -- there is an article in U.S. News & World Report this week in which I am prominent as Chicken Little or Cassandra. And even in that article, there is no one in the article -- the people in opposition -- no one is saying, Well, the capacity is X and the demand is Y, and therefore is it okay.

They all say something to the effect, "we believe Metcalfe is Chicken Little." Or, "he is a clueless, opinionated jerk," is another form of the counter-argument.


This is known as an ad hominem attack. And the fact that I am getting them means, aha, I am on to something here.


So I am waiting for facts to come flying back, to show that these projected collapses are not going to occur.

The "Available Excess Capacity" Argument

You often hear the dark fiber argument, that there are more optical fibers installed in the world that are just lying unused -- excess capacity -- it is so cheap that, to meter it, would cost more than to provide it. Have you heard that argument?

Well, you just did.

But that overlooks the fact that the optical fibers are a very inexpensive part of the system. There is the transmission equipment, the switching equipment. And then you go a little further up, through the servers. And there is a lot to be paid for, not just the fibers.

And some of the capacity problems we are having on the Internet today are in the routers. The routers just do not switch fast enough. And they lose packets, because they don't have enough memory and the data comes in too fast and they just throw the packets away in desperation.

Expansion of Computer Capability compared to Expansion in Bandwidth

In the early days of Ethernet, we used to argue that -- people used to argue that the cable costs 12 cents a foot, whereas twisted pair for telephone cable costs 2 cents -- I forget the numbers, but 2 cents a foot. Both of them cost $10 a foot to install. So the cost of the actual cable was irrelevant.

So let me talk about a few laws here. I would like to talk about four laws -- Grosch's Law, Moore's Law -- and Metcalfe's Law, of course, which is a part of the new paradigm -- and Grove's Law, which I just learned yesterday from Gordon Moore. You know, I am sitting next to Moore's Law, I might as well ask him about it.

Herb Grosch said that the cost of a computer grows as the square root of its speed, which many of us have always interpreted as a defense in the mainframe days that larger computers were better. I have talked to Dr. Grosch recently, and he does not disagree with that. But he also says his law was about consuming of computers by software people. That is, he was an early observer of the "Grove giveth and Gates taketh away" principle, where the software people consume the processing power faster than it can be made.

Then Moore's Law says that density of semiconductor devices grows -- doubles every 18 months, and of course, Dr. Moore has revised that law several times since he first stated it in 1965. He now believes that people refer to Moore's Law as anything in semiconductors that when planted on semilog paper results in a straight line.

Metcalfe's Law, which I have already mentioned -- the value of a network grows as the square -- and now Grove's Law. Grove's Law: Bandwidth doubles every hundred years.


And what he is referring to, I think, is the now grossly obvious difference between the performance of the fiercely competitive computer industry and the telephone monopolies that were established some 60, 70, 80 years ago. And the telco monopolies have been under-performing for a long time, and now it is really sickeningly obvious. And most of the people that I talk to would prefer to have a doubling of bandwidth over a doubling of processor speed.

That is, they've got all the processor speed they need, with a few counter-examples. What they want is more bandwidth. And the reason they are not getting it -- aside from the fact that computer networking is really sophisticated and it takes real geniuses to make any advances in it -- is the telco monopolies standing in the way.

Significance of Deregulation

So we have a new law, the Telecommunications Deregulation Act of 1996, which is celebrated widely, but should not be. Customer telephone equipment was de-monopolized in the seventies, I think. The Carter phone decision was '68 or something like that. And that allowed the growth of things like Hayes modems. That is, it used to be true that the telco monopolies controlled devices that you connected to the telephone network. That was eliminated.

It was eliminated under the fear-mongering of the time: "If you ever did that, the whole telephone network would fall apart, because the telco's needed to control that interface, and those devices would break the telephone network." Well, that turned out to be nonsense. And CPE, as they call it, has boomed since.

Then there was the long distance market, which was deregulated. So now we have MCI and AT&T and Sprint and a growing list of others competing fiercely in long distance, and long distance rates going down.

And now, we were supposed to, this time, deregulate local telephony -- that last mile -- only we failed. And this is only half serious, but the telco lobbies invented this thing called the Communications Decency Act, which they stuck on to this bill, the Telecommunications Act, to keep all of us busy, arguing about freedom of speech and freedom of expression and censorship and various amendments to various constitutions, all the while the really important part of this bill was being engineered by the telco lobbyists to defend the local monopolies.

And the way you know that this is true is, right after this bill was passed, there was wild celebration. And shortly thereafter, U.S. West announced its intention to buy Continental Cable, and Pacific Bell and SBC decided to merge -- or proposed to merge, and now Nynex and Bell Atlantic are proposing to merge. And our only hope is that those mergers will be disallowed by the Justice Department.

Those mergers do not bring competition to the local loop. They are unnecessary accumulations of market power, and they will continue making Grove's Law true, which is that bandwidth doubles every hundred years.

Forms of Competition

QUESTION: Isn't there competition between telephone companies and cable TV companies?

MR. METCALFE: And there is satellite, too, and there is cellular, too. And I think our goal should be to set those people at each other's throats, competing fiercely, the way Silicon Valley works. I mean if we could get Continental Cable or TCI competing with Pacific Bell the way Intel competes with National and so on, that would be great. And so cable modems are a promising alternative to ISDN.

There is a problem with cable modems, which is that they do not work.

QUESTION: They are expensive too, aren't they?

MR. METCALFE: I think not working is a more serious complaint.


But, you are right. They are not only expensive, they do not work. But they are going to work eventually, and we should be delighted that those people try as hard as they can. What we do not want is the telephone companies to buy them up. Because then the competitive energy will be drained from the system, and we will go back, as I said, to Grove's Law.

The difficulty there is that the whole television industry is broadcast-oriented, and it is analog, it is one-way, it is advertising financed, it is a completely different business model, completely different technology from what the Internet needs. And it can be changed. It is going to be much harder than we are now hearing.

In fact, we heard a lot was going to happen last year that did not. It is an eventual thing, and we should be rooting them on.

Two-way Television

QUESTION: How does this all relate to two-way television?

MR. METCALFE: It is much easier to broadcast than it is to go two-way. And we are forever hearing that wonderful progress is being made in converting the TV network to be two-way. And there is progress being made, it is just it tends to be exaggerated. Frankly, I think more promise would be in some sort of hybrid network, where you use, say, satellites of cable down and some sort of cellular or telephone up -- maybe that would work sooner, better. But there are many, many combinations, and we should get them all competing crazily.

Historical Parallels

QUESTION: Aren't there parallels between this situation and what happened with other industries?

MR. METCALFE: There probably are lots of parallels. I know the National Center for Research Initiatives, which is behind a lot of the Internet work that is going on these days, on the side is doing a series of infrastructure studies. And I have read the one on railroads. And I think they have one on telephones. And maybe they are going to do electric power soon.

They write these little books -- you should get them -- they are very small, and they sort of do exactly what you do, which is try to develop the parallels. Of course, unfortunately, the books say here are the parallels and here are the non-parallels, and it leaves it to you to -- well, to all of us -- to try to sort out what we can learn from that.

Data Compression

QUESTION: How important are advances in data compression?

MR. METCALFE: That is one of the big swing factors. I get the sense that compression is well understood and pretty far along, and the extreme claims you hear are probably over the top. I have run out of anything to say about that.

Compression is very important. There is a lot going on. This is probably how we will prolong the life of the twisted pair lines. When we want to go even faster, we will just compress things.

But I have compression in all my stuff, and I don't notice it helping me. I have, allegedly, a 4-to-1 compression scheme in my ISDN bridge into the Internet, and I have never seen 500 kilobits per second. I have seen 200 bytes per second. I have seen -- I typically get 40 or 50 kilobits per second through my 56 KB line. Now, if the compression was doing me a lot of good, I would see much more than that.

Maybe it is just being held back by slow servers. So, yes, compression is great. I look forward to getting something compressed someday.


I know. It is just my files. Everyone else's files are compressed. Just mine are not compressible.

Transition to Metered Usage

So the next generation Internet will have metered usage. I like to take this -- to really drive the intelligentia crazy, I like to advocate that we should charge per gigapacket meter. A good metric for charging would be how much data you carry, -- and I've chosen gigapacket meters as a useful unit there.

Because, frankly, it does cost more to carry data from Europe than it does from -- what's near? -- Washington. It does actually cost more. And it takes longer and it uses more resources. And it goes through more hops. And why not charge for that?

Need for Quality of Service Guarantees

Guarantees of service. Quality of service guarantees. There are none in today's Internet. There should be some. Of course, there are no guarantees ultimately, but recourse, I guess, would be, in the Internet, there should be no more than 30 hops -- I made that number up -- between you and your most frequently visited destination. That might be something you might ask of your ISP.

Because one of the phenomena of the new Internet is the number of hops between sources and destinations is growing. And there are arguments that this does not matter. But I know that it matters. Because I know that every hop is a source of delay. It is a source of risk. It is a source of congestion.

And the reason that the hops -- my theory about why the number of hops is going up is that there is no economic advantage in providing backbones for other people. If you're a NAP, what is your incentive to provide -- you do it, because you are part of the new paradigm, but, after that, why do you do it? They are not going to pay you for it.

So, what happens is, instead of everyone going up in sort of a low hop configuration of the Internet, with a well-organized hierarchy, you sort of connect horizontally to agreeable people next to you. A very good example of this is in rural Maine where I live, I have a 56 KB line to a guy in Owl's Head, Maine, who has a T-1 line to a provider in Augusta, Maine, who has a link into SprintNet.

Worse than that, I have put the local K-8 school -- I bought an extra router, in a little deal I did with the ISP -- that school's router is connected through a four-mile link into my router, which goes to this ISP. And he does not charge me extra for this, because he is public spirited. So that school goes from there to me to him to Augusta before it gets to SprintNet. And who knows what happens there.

Anyway, the measurements of the diameter of the Internet are growing. Let's say, if you know you are frequently going to be communicating someplace, maybe you would like your ISP to assure you that there will be no more than 800 hops or 30 hops or whatever the right number of hops is.

Need for Encryption

There will be a settlements mechanism, as I have kept harping. Encryption will become no longer the exception. We will clear up the federal laws against strong encryption -- get rid of them. Install encryption standards.

But we will go one step further, which is we will now -- we will all take the moral position that all of us should send everything that we send encrypted. So that the people who have a reason to send information encrypted will not stand out.

I heard this argument recently. I loved it. Remember the Danish king wore a yellow star during World War II to confuse the Germans about who was Jewish and who wasn't. It's the same thing. We should just all encrypt, and that way, anyone who has a secret will not stand out.

Need for Better Audio and Video Capability

Another problem is audio and video. The Internet should carry audio and video. In its current configuration, it will not, in any appreciable scale -- and you should not welcome the news of real audio and real video, because if you do not like the performance of the Internet now, you are not going to like it after everyone is on.

I understand this particular talk is going out over M-Bone. And I would like to apologize to the rest of the Internet for that.

But what is needed is a reconciliation of the audio/video technology -- roughly called ATM -- and IP. That is, the Internet and the technologies which have been developed to carry audio and video need a very smooth reconciliation. I think that is actually happening. But there are two intelligentsias there. One emanating from the telephone companies, the ATM people, and one from the IATF. And they do not like each other. And there is hostility there.

But I think the next generation Internet will have an accommodation -- a merging of those two worlds so that we can carry audio and video in quantity, in scale on the network.

Need for Zoning Restrictions

And now, just in case I haven't upset you enough, I think the next generation Internet will have zoning laws. And this relates to the Communications Decency Act and other things. Which is, I was noticing what a lovely place this campus is. And I bet you there are rules about what you can hang around here. Like billboards in the middle of the lawn probably would not be welcome.

Is that right? If someone put a billboard in the middle of the lawn, would the university remove the billboard, do you think?

Now, the removal of that billboard would probably constitute, in some people's mind, a violation of the First Amendment. In my mind, removing the billboard would be consistent with the rest of the Constitution -- you know, pursuit of happiness and quality of life and all of life, liberty and the pursuit of happiness.

The right of freedom of speech is matched by my freedom not to listen. Anyway, what this zoning thought is, is that the Internet will need to have zones, where children can be healthy and free, and where religious communities can grown in whatever way they want. And they can all live on the Internet without sort of dumping on each other all the time and forcing themselves on each other.

But I think we are going to have zoning on the Internet. The Communications Decency Act is an awkward attempt at that. And it needs fixing. But it is an attempt at zoning. It is an attempt to sort of keep the Internet civil and a place where you might want to let your children go.

Expanding Internet Access

QUESTION: How do we expand access to the internet to the non-business community, through taxes?

MR. METCALFE: It is a whole, separate, complicated subject. I am against taxes in general. Government, in general, can all go to hell for all I care. But I thought I would let a little ideology leak out there.


So let us assume that having access to the Internet from schools and libraries and other useful societal institutions is valuable. I share that with you, I guess, from your question.

The question is: How do you accomplish it?

Two basic methods. One of them is to make it a prerequisite of being a supplier in this market, that you give resources to those worthy institutions as a quid pro quo of doing business in this market. That is the traditional method. That is what I consider to be the bad method of solving this problem. Because it distorts -- first of all, it justifies monopoly and it increases regulation, and it distorts the competitive environment. Although it has a good intention, which is to provide these services to worthy organizations.

The other method is the one where we let a competitive market flourish for providing these services, and we give the money to buy those services to the worthy institutions, and have them buy them on the competitive market. So that those worthy causes have the benefit of a competitive scenario.

And you will see the dichotomy I am trying to draw there.

Improved Educational Use of the Internet

QUESTION: How can we achieve the goal of better educational use of the Internet?

MR. METCALFE: Well, I am trying to separate, too, the goal from the method. The goal would be that everyone should have access to quality education, which, under the current system, is a rapidly declining opportunity.

So how would you propose to do it differently?

I would propose to do it differently by eliminating the public school monopoly and increasing choice among the people who are customers of the schools. And that includes everybody would have choice, not just -- my kids can go to any school they want in the world. They have choice. I am worried about the kids who do not have choice. Give them a choice, too, by giving them a way to buy the school they want.

And then give the schools a choice as to what textbooks they buy, instead of having that centralized. And break all monopolies. Whatever the question is, my answer is eliminate monopolies and give people choice, have competition.

I am glad we settled that matter.

Instituting "Micro-Money" Transactions on the Internet

Another aspect of the next-generation Internet would be micro-money, a system for supporting transactions below the penny level, so that you can economically buy something for a tenth of a penny or a hundredth of a penny or a thousandth, a mil.

Now, right now, we do not have that. It is very uneconomical. Even the telephone companies, who are quite good at -- they charge for 6-second intervals, as I recall -- even that is too granular. So we, through the Internet, which has the technical capability, we should institute a micro-money system, so you can afford to buy things in very small pieces.

And the theory is that if you had such a thing, then you could lubricate the economy that is being created on the Internet. So that many more things can be bought by many more people. So, for example, you could pay for gigapacket meters. Most people who object to metering actually are objecting to paying a lot of money. It is possible that you could meter and pay very little money. And the micro-money system would be progress in that direction, where there would be metering, but you would be paying so little that you would not have reason to complain.

To me, the big upside, though, in micro-money is a completely new market for intellectual property. My column is the closest example. Right now, my column is delivered to its relatively few readers through advertising support. And I am lucky because my readers happen to be the kind of people who attract advertising support. My readers buy things -- lots of things. So, therefore, it is easy for me to find advertisers who want to pay for my newspaper to send to those people.

But sometimes in my egomaniacal moments, I would like to write for broader audiences than that, like people who do not buy a lot of things. And then, I would like to be able to charge for my column. So, if you want to read my column, you just send me what? If the price is $10, I would not get many takers. I offered my readers 2 cents -- there was resistance at that level.


So I need a micro-money system to broaden the readership of my column.


There are still people who will -- you know, even -- I have actually said, Well, you know, you could buy the column for 2 mils, and there were still people who say, Well, what if I didn't like? I said, Well, for 2 mils, that would just be the risk you would have to take.


So, micro-money. Now, micro-money is a technological problem. It is a standards problem. It is a political problem. It is a big problem. But, fortunately, it looks like we are going to have many kinds of money. What I am advocating here is in the next-generation Internet, in addition to having regular money, like Visa cards, we should be careful that we end up with a micro-money system, so we can buy very small things. This will solve a lot of the intellectual property problems that we perceive we have. It will deal with the have-nots, these people that we worry about so frequently in discussions like this.

Making things cheap must be helpful to have-nots, right. But if you limited how cheap you can make them by your money -- that is, if you can only afford to have expensive things, that slows progress down.


Another characteristic of the next-generation Internet is advertising. The American public, in particular -- it is a worldwide phenomenon, but I am certain, in America, we are used to not paying for a lot of our information, and that is not going to change suddenly. We are used to having it paid for us by advertisers.

I have heard estimates ranging as high as 75 percent of the information we now enjoy, including television, radio, newspapers, and magazines, is paid for by advertisers. That will be true of the Internet. That is, a substantial part of the Internet is going to be supported by advertising.

You know the current debate about whether Internet terminals at $500 are a viable? There is a big argument about whether the $500 Internet terminal will be an attractive proposition compared to, say, a $1,000 personal computer. Well, how about a 0-dollar Internet terminal, where instead of buying a box called a PC, you buy a service called the Internet, and the box does not cost you anything, you just pay for the service?

Zero is pretty cheap. And that would be paid for by advertising. The next-generation Internet will be hospitable to advertising. Not spamming, not the kinds of advertising that do not work, but the advertising industry is pretty sophisticated in figuring out how to only annoy a small fraction of the people most of the time.

And implicit in that will be a measurement system, which will have to be developed so you can measure the effectiveness of advertising. And, of course, that has to be done carefully, because it brushes up against junk mail, privacy of information issues. But I think that is inevitable. That will be in the next-generation Internet.

Collaborative Filtering on the Internet

Another thing will be collaborative filtering, which, in some ways, is an alternative to advertising, and it is an alternative to journalism. It is a way in which a community of people help each other decide what is good and what is not, by sharing each other's experiences, collaborative filtering.

This is a particular technology that I have sort of decided is going to be very important For example, CD-ROM's--you see one, you are asked about it, you express your opinion about it, that opinion goes into a database, the database then collects the opinions of millions of people. And then, when you need an opinion about a CD-ROM, you can find a good one and you do not have to rely on Siskel and Ebert, or whoever the CD-ROM equivalent is at that.

The Role of Editors and Publishers on the Internet

One last comment on next-generation Internet. Writing, editing and publishing are often viewed as intermediaries that need to be eliminated. You will often hear this said in Internet, "Oh, thank goodness we have got the Internet now, we will not need editors and we will not need publishers, because we can all now talk directly to each other."

Well, that does not scale up. And I think we are going to have a resurgence of writing. Writing is already being encouraged by the Internet, and I have met many English professors who are delighted at all the writing. They do not particularly approve of the quality of the writing, but at least people are writing again.

And now editing will be important again, and publishing. Although there is a new twist. In the publishing business, there is a dichotomy between church and state, between advertising and editorial. And there are all sorts of elaborate traditions and rules about how you keep advertising and editorial separate.

I have just gotten into this a few years ago, and begun to learn the disciplines. And some publications are more disciplined than others in how they keep advertising, which is done principally in the interest of the advertiser, separate from editorial, which is done principally in the interest of the reader. And when you begin to confuse those interests, the publishing model breaks down.

Emergency of more Community Content in the Internet

There is a third kind of content now -- important content. In the old days, it was called "letters to the editor," but it is the content in your publication which is not generated by advertisers and it is not generated by the editors. It is generated by the readers, by the community. And there have to be traditions developed now about the rules separating church, state and community.

When is it right for advertisers to send provocateurs into a community forum and pretend that they are participants? And when do you get to edit community input before you put it in -- is it proper to edit it? How do you edit it? And so on. All the rules for the separation of church, state and community.

Institutions hurt by the Internet

QUESTION: What institutions will be hurt by the Internet?

MR. METCALFE: Well, the Internet threatens many institutions, and universities are among those, especially placed-based universities. I guess most universities are very place-based. The University of Virginia is extremely place-based. And universities are threatened. But I think you are sort of close to my sense of what the role of universities is. It has to do with sorting through the garbage and choosing the good stuff. But how that works itself out, I have no idea yet.

The Role of Government

QUESTION: Shouldn't the government restrict its role to anti-trust regulation?

MR. METCALFE: I agree entirely with the premise of your question, which is that free markets are going to generally work things out, and government's role should be minimal. But in that minimal is, I would post antitrust as one of the things that government needs to do, because of the advantages of incumbency.

So, even though I am something of a right-wing kook, I will advocate -- I have advocated that the Justice Department should stop Microsoft from buying Intuit -- if you remember that episode. Because that, to me, represented a case where the advantages of monopoly were about to be exploited to the detriment of competition and choice, and that that was inappropriate. And right now, with the Telecommunication Act of '96, we need a de-monopolization act of '97 or we need the Justice Department to step in and fix the telco's, whose monopoly is oppressive and holding everything back.

So the proper role of government is much lower than it is, but there are proper roles. I would list antitrust high among them. Financing of university research and networking would be high in that. I said that just to win you over.


We often point at the ARPA-Net and the Internet as success examples of government support of research. And I think that is a good example. But governments tend to overdo things. In other words, they never stop at the rational point. The bureaucracy is transformed. It goes from a mission orientation to a jobs program suddenly. And so now the government is really itching to crawl all over the Internet and tax the hell out of it and control it.

The CDA, the Communications Decency Act, is a clumsy effort for the government to get involved. I think they should get involved in some degree -- zoning, you know, creating some zoning ordinances within which the free market can operate.

Big government and big business are equally prone to complacency and coercion. That is why the monopoly of government and the monopoly of big business are both bad. Because when you have a monopoly of any kind -- the thing that will keep the ISP's from -- what was that word? -- voracious practices-- I like rapacious, that is how I used to describe Microsoft -- not that it is not rapacious anymore, it is just that I have gotten tired of calling it rapacious.


When you are big and your customers have no choice, it is a natural human process to become complacent. And then, when things do not go your way, your complacency turns to coercion. And that is why big businesses and big governments should be made small again, both of them.

Well, we have 2,000 ISP's. If the Internet goes the way it is going, those 2,000 ISP's are going to be gone soon, and you are going to get stuck with the telco monopolies again. That is, they will use their monopolies to kill the -- right now, there is choice in ISP's. I can switch. Even in rural Maine, I have a choice of three ISP's, and my local telco is threatening to become an ISP -- a very small company. And Nynex and AT&T are imminently offering. So that is good.

And the bad part is if we do not keep the monopolies from killing their competition. I sort of agree with the premise that -- what I detect is the premise of that question, although you said "big business" with such a sneer. Big business is no worse than big government. Big government is worse, because big government has F-15's.


The only reason I pay my taxes -- I pay income tax. I pay a lot of income taxes. And the only reason I pay them is not because I admire Thomas Jefferson, it is because if I do not pay them, they are going to send people with guns to my house and take the money. So I pay.

My particular crankiness you have detected already, and it has to do with an abhorrence of big, monopolistic controlling things, including government and big business.

The Long-Term Impacts of the Internet

Now, the long-term impacts of the Internet. The Internet will contribute to the melting pot, the world melting -- not just the U.S. melting pot, the world melting pot, and hopefully will eliminate a lot of the divisions. In recent years, we have been celebrating our differences. And I think that has been a mistake, in that we should be celebrating the melting pot and the removal of divisions which cause us periodically to want to kill each other.

And through the connecting of people through the Internet, perhaps a lot of the differences that lead to violence will melt away. This includes -- Seymour Cray is interested in the decline of nationalism, which would be a form of this. And he sees the Internet as a way of doing that, and I had to agree.

I think the Internet will make markets efficient -- more efficient than they are today. We were lamenting this yesterday in our conversation. If you have an Internet where everyone can communicate with each other, that will turn everything into a commodity, and make everything cheap. And, "Where will the profits be?" was one of the questions.

I took that as good news -- progress. I mean I do not think the profits will go away, but they will be chased around and minimized and new business forms will rise up. But efficient markets are one of the things that is promised.

Increases in personal privacy. The current consensus is that the Internet threatens personal privacy. I think not. I think the Internet will improve personal privacy, because it comes with a set of technologies that can improve the way we protect what is private to us from others.

Every time I give my credit card to this person who I would probably not invite to my house, and they disappear around the back of the restaurant somewhere, and they emerge 10 minutes later, what happened back there? Can I follow you around? Where did you take my credit card?

And I think the technologies of the Internet promise to improve personal privacy. They are not so much a threat.

The rise of intellectual property or the information economy, where people will be able to buy things for 2 cents or 2 mils. But, in general, the Internet promises dramatic improvements in freedom and prosperity. So that is code word for I am an optimist about how the Internet is going to go.

So Luddites and socialists and terrorists, criminals, lawyers and bureaucrats, beware, the Internet is here.

Solving Internet Problems

QUESTION: How do we solve the problems with the Internet without excessive government intervention?

MR. METCALFE: This is a hard question, and I do not have the answers, but I know that the alternative lies in this direction rather than in apparent anarchy, which is causing the Internet to decline. I have seen it happen before. The way I am trying to fix the problem is to convince my readers that they should demand more from their ISP's. I have faith in buyers.

The making of the Ethernet standard, which was important in the evolution of my career -- I watched the vendors meet to try to make a LAN standard. And a LAN standard eventually resulted, and it was a horrible, dirty, terrible thing, probably much like what goes on in Washington every day.

But what made it work in the end is customers showed up and said, how are you guys doing here with that LAN standard, how is it going? And confronted with their customers, the vendors caved. Because they could not --you know, these are people that they sell to -- they could not justify the pettiness and narrow parochialism that they had been bringing to the standards process. But when those customers showed up, they got cooperative and a standard was made. And it was not made by the federal government, it was made by the IEEE or a subgroup of the IEEE.

So it is that kind of thing, leaving out a lot of details, that I am hoping from NANOG and the international -- it has to be the international version of NANOG.

We are going to see that this summer, on July 19th, when the Olympics start. And the Olympic Committee is giving these incredibly wonderful speeches about how the Internet is going to be a way for the entire world to come and learn about the Olympics.

Now, the Olympics is an international event, and a lot of the people interested in it might come to the Internet from the international channel. Well, if you think the U.S. Internet is bogged down, you should see what happens when you go across the Atlantic. So there are going to be bad things happening. "Beware of the ides of July" is the headline of that column, when the international tries to come to the Olympics.


So, just pursuing your question further, so the organization needs to be a GII, a global Internet management, standard-setting -- maybe the IETF could expand its activity -- oh, God forbid, I am suggesting this -- but, anyway, they should expand their activities to include some semblance of a management process, tracking problems, reporting them.

Right now we have a reporting problem, because the commercial interests involved are reluctant to report their failures. When the network goes down, very few of them announce this or record it anywhere. They do not want -- why upset people? They did not notice, why tell them?

The FCC has this problem with telephone companies. The Nuclear Regulatory Commission has this problem with nuclear power plants. If a nuclear power plant reports some difficulty, the NRC is set up to just punish the hell out of them for it. So they do not. They are reluctant to report problems. Well, this is not helpful when your nuclear power plants are sort of suppressing their problems because they do not want to attract attention.

Well, the same thing is going on with the ISP's. They do not like reporting their problems, but they should be encouraged to. And I think they could be encouraged to if customers made them. So I am trying to organize my readers to just demand reasonable things from their ISP's, and create a more systematic, organized -- with a bright light on it -- a bright light so that we can see what is going on.

Back to Index

Copyright © 1997 The Computerworld Smithsonian Program All rights reserved.