Making Some Sense of the Madness

The Natural Evolution of the Internet Industry

Originally published 4/1/96, revised 8/21/96
by Sky Dayton,
Founder & Chairman,
EarthLink Network, Inc.

State of the Internet industry
Only two years ago, the Internet was virtually unknown in the mass-consumer market. Today, it is a phenomenon reshaping global communications for millions of people. Almost every industry is touched by the Internet's fanatical momentum. It is growing faster than any other new technology industry in history, and it has the potential to change the way people interact on Earth.

In many ways, the Internet's communications properties have acted as a catalyst to its own growth, making it a self-fueling entity.

Despite its growth and its assimilation of whole industries, the Internet is still a nascent market. Many in the industry fear a backlash if the Internet fails to deliver on its hyped potential fast enough. Even now, media pundits have begun to circle overhead looking for the opportunity to strike and be the first to predict such a backlash.

Most agree the Internet will be the foundation of what some call the "Information Superhighway", a communications system that will tie together all existing voice and data applications on an open, inexpensive and fast network. However, the current Internet industry lacks a clear definition of how it will mature toward this vision.

As in all markets, maturity will bring segmentation of products and services within the Internet industry. An understanding and prediction of this segmentation is needed.

Need for definition in the industry
We need a segmentation map of the Internet industry. Without some sort of guideline of how the market will mature, companies will have difficulty putting their potential products and services in context with other companies and in focus within the rest of the industry.

Many large companies today are pursuing unrefined "Internet strategies". Thousands of start-ups are also clamoring to produce products and services for the Internet. Until recently, investors hoping to cash in on the craze were buying any stock with the word "Internet" attached to it, and analysts continue to have difficulty differentiating thousands of new product and service offerings.

Additionally, industry experts are still trying to explain the failures of dozens of previous ventures in the interactive video and proprietary online services businesses, especially those backed by top-tier players such as AT&T, IBM and others.

Companies that don't predict segmentation will have their offerings segmented by market forces nonetheless. Hundreds of companies, from start-ups to blue chips, went through this experience in the PC revolution.

This paper gives participants in this new industry a guide to the coming segmentation, thus allowing them to better define and position their offerings in the context of one another.

The premise of this paper is that Internet market segmentation will be based on the structure of Internet networks themselves.

The OSI model
The Internet marketplace is already segmented in a way analogous to the basic architecture of networking. Network architecture is best defined in the Open Systems Interconnection (OSI) model developed by the International Standards Organization (ISO) over the past 15 years.

The OSI model was created to provide an open-standard reference for network development. It lets network developers work on different parts of the same networking problems in a way that will allow their final work to be compatible with other developers' work.

The OSI model describes the building blocks of all networks. It lays out how a network is built, from the foundation up. Computer networking began as a wire connecting two computers. In order to communicate, the voltage on the wire was modulated. As networking evolved over the years, engineers began building up from this foundation, one layer at a time, with each successive layer allowing greater network functionality.

Much of today's networking technology, including the foundation protocols of the Internet, TCP and IP, was developed without direct reference to the OSI model. So were hundreds of other protocols. However, OSI is an attempt to put networking protocols in context.

Understanding the OSI model
Here is the OSI model:

Layer	Name	Description	Example
7	Application	The communications applications themselves.	email, file transfer, client/server applications. Applications such as Netscape & Eudora talk to this layer.
6	Presentation	Syntax for data conversion, makes session layer available to application layer.	ASCII, binary conversion, encryption and decryption, sockets
5	Session	Starts, stops and governs transmission order.	Sockets, synchronization
4	Transport	Ensures delivery of the completed message.	TCP, SNA, UDP
3	Network	Routes data to different networks. Forms packets.	IP, x.25, IPX, AppleTalk, Routing
2	Data Link	Transmits from node to node. Divides bits into frames.	Ethernet, Token ring, Frame Relay, Bridging
1	Physical	The connection medium. The hard, physical connection.	RTS, CTS, RS-232, copper, fiber, wireless.

The access sector of the Internet market is currently populated by Internet access, cable and telephone providers. Each company in the access sector can be plotted on the OSI model above.

The company I founded, EarthLink Network, is an Internet access provider focused on the consumer market in North America. For most of our network, we purchase layers 1 through 4 from our partner, UUNET Technologies. They deliver TCP/IP packets (layers 3 and 4) to our customers over dial-up modem lines in hundreds of cities. They, in turn, purchase layers 1 and 2 from local telephone companies. These companies deliver dial-tone or other layer 1 and 2 services to UUNET.

If UUNET can send a ping (a TCP/IP packet from one point to another) to our customer, their job is done - the customer has an Internet connection. We take over at that point and deliver the services the customer uses to communicate on the Internet: email, Usenet, ftp, Web, etc. These services operate on layers 5 to 7.

Early analysts might have seen our relationship with UUNET as competitive. EarthLink predicted industry segmentation, however, and recognized the relationship was complimentary. UUNET decided to focus its business on layers 3 and 4 principally. It buys layers 1 and 2 from carriers and packages the whole thing up into a 4-layer product.

Other Internet access providers, such as Netcom and PSINet, have attempted to build 5 layers at once, all during massive growth. While this strategy made sense early on, these companies will have to defend many more layers against competition. Their competitors will specialize more and more on one or two layers, resulting in better efficiencies than Netcom or PSINet could ever have on a multi-layered strategy.

Since the first edition of this paper was published, PSINet has refocused their business on their traditional core competence: layers 3 and 4. After several quarters of diluted attention and stunted growth, they sold off their consumer dial-up access business and concentrated on business-to-business and wholesale Internet access, where they have always been strongest. With this new, focused strategy, they will be able to compete much more effectively in their market segment.

Other companies have failed to embrace market segmentation by assuming their expertise in lower layers would translate into an advantage in higher layers, particularly where their implementation required them to provide all 7 layers themselves. Today, this is tantamount to IBM's early attempts to own all aspects of the personal computer business.

Recent examples of such disasters were AT&T's foray into computer hardware and online services, USWest's interactive video services and MCI's consumer Internet service. These companies' traditional technology expertise did not translate into expertise on higher layers.

Though telephone and cable companies are pursuing Internet access strategies today, their core business is on layers 1 and 2. With recent deregulation, it is detrimental for them to leave layers 1 and 2 unguarded from major new competition to pursue a completely new business on higher layers.

Layers 1 and 2 are the foundation of the Internet. Because telephone and cable companies were regulated monopolies for so long, they have let this foundation languish. They have not kept up with sweeping innovations in the upper layers. As a result, most consumers now suffer with an Internet experience at a mere 28.8k/bits per second; their computers can process bits thousands of times faster than the network can send them.

The telephone and cable operators have a significant task before them to upgrade and maintain the foundation of the Internet in a newly competitive market. In the long term, they will be more successful focusing on new challenges and opportunities in their traditional lower-layer businesses than pursuing purported opportunities on the upper layers.

OSI model applied to all Internet industry sectors
The OSI model defines how the technology of the Internet underlies the segmentation of its products and services. The industry can also be subdivided into sectors which classify the different types of products and services available on these network layers.

The Internet currently has six market sectors:

1. Hardware
2. Software
3. Access
4. Content
5. Services
6. Expertise

In order to see how companies in each of these sectors operate in the context of one another, all can be classified on the OSI model. The chart on the following page is a random sampling of companies in each sector.

Name	Hardware	Software	Access	Content	Services	Expertise
Application	Apple, Intel, Dell, Gateway	Netscape, Microsoft, Quarterdeck	EarthLink, Netcom, Sprynet, GNN	AOL, Compusrve, Prodigy, CNN, WSJ	Yahoo, Excite, Infoseek, Lycos	CKS, Digital Planet, USWeb
Presentation	Sun, SGI, HP, IBM, DEC, Intel	Netscape, Microsoft, Interworld	EarthLink, Netcom, Sprynet, GNN, Best	AOL, Compusrve, Prodigy, BBN	CyberCash, RSA, Compusrve, BBN, I/Pro, Netcount	CKS, Digital Planet, USWeb
Session	Sun, SGI, HP, IBM, DEC, Intel	Netscape, Microsoft, Interworld, Xing, RealAudio	EarthLink, Netcom, Sprynet, GNN, Best	AOL, Compusrve, Prodigy, BBN	BBN, Compusrve, I/Pro, Netcount	CKS, Digital Planet, USWeb
Transport	Cisco, Bay, 3Com, Wellfleet	FTP, Netmanage, Network TeleSystems	Netcom, UUNET, PSINet, Concentric, MCI, BBN, Sprintlink	AOLNet (ANS), Compusrve, Sprintnet, BBN	BBN, Compusrve, Network Solutions	Anderson, EDS, Perot, BBN
Network	Cisco, Bay, 3Com, Wellfleet	FTP, Netmanage, Network TeleSystems	Netcom, UUNET, PSINet, Concentric, MCI, BBN, Sprintlink	AOLNet (ANS), Compusrve, Sprintnet, BBN	BBN, Compusrve, Network Solutions	Anderson, EDS, Perot, BBN
Data Link	USR, Ascend, Cascade, Stratacom, AT&T	NorTel, AT&T	MCI, TCI, Sprint, Worldcom, AT&T	MCI, TCI, Sprint, Worldcom, AT&T	MCI, TCI, Sprint, Worldcom, AT&T	Anderson, EDS, @Home, AT&T
Physical	USR, Ascend, Cascade, Stratacom, AT&T	NorTel, AT&T	MCI, TCI, Sprint, Worldcom, AT&T	MCI, TCI, Sprint, Worldcom, AT&T	MCI, TCI, Sprint, Worldcom, AT&T	Anderson, EDS, @Home, AT&T

Many companies overlap onto different layers in more than one sector. Most companies, however, generally maintain their core business on one or two layers in one sector only.

The OSI model specifies that each layer only deals with the layers immediately above and below it. Similarly, companies will be most successful doing business with other companies on the layers immediately above and below them. As an illustration of this, we have seen many recent failed partnerships between cable or telephone companies and computer software and hardware companies. These companies are many layers apart, as the OSI model shows. The best partnerships are between companies occupying neighboring layers.

Also, complications in developing and deploying networks and network-based products are directly proportional to the number of layers involved. Conversely, the fewer the layers involved, the simpler and more reliable the networking technology.

With more layers involved, the technology becomes more complex, but it also allows much greater flexibility. For example, Morse code operated at the first layer only, and telephone technology deals only with the first and second layers. Both afford little room for innovation compared to the seven-layer full Internet connection.

Because Internet technology is so versatile, it will be difficult to fit the industry into a "commodity" description. While one or more layers may become commodity businesses, there is so much flexibility in the implementation of multiple layers, it will be a long time before the entire industry commoditizes, if ever.

Summary
The Internet has the potential to become the greatest communications medium in history. It will take thousands of developers at hundreds of companies to provide the innovations necessary for the Internet to achieve this potential. Whether or not these companies become a financial success depends largely on their anticipation of the market's segmentation.

Many companies will not foresee this segmentation and will fail to position themselves to occupy a layer, or segment, of the industry. Or they may pursue a segment but fail to understand its relationship with the layers immediately above and below.

Companies that understand and embrace the natural segmentation of the marketplace will be able to make wise decisions about partners and competitors. By anticipating changes in the market, these companies will be principal contributors to the Internet of the future.

What's all this about the Internet "collapsing"?

by Sky Dayton,
Founder and Chairman, EarthLink Network, Inc.
8/26/96

You may have seen recent magazine and newspaper articles predicting the imminent "collapse" of the Internet due to excessive load and other problems. These articles cite recent outages at Netcom and AOL as proof. Bob Metcalfe, the inventor of ethernet and now an industry columnist, has been prophesying a catastrophic collapse of the Internet in 1996. For some reason, Metcalfe is doing everything he can to promote his prediction, short of renting the Good Year blimp. Luckily, these "doomsday" predictions have little to do with the real state of the Internet.

What's really going on
The Internet, of course, is like a system of roads and bridges, all interconnected. Once you're on a particular road, you have access, via highways and other connection points, to any other road in the system.

Today, there are traffic jams at some points on the Internet's system of roads and bridges. But the statement, "The Internet is overloaded," is about as ridiculous as "America's roads are overloaded." Which roads? Where exactly?

Because the Internet is so little understood (especially by the press), traffic jams in particular places are taken as a generality for the entire network. The specific points where the network is jammed are not discussed much. Instead, the problem is being generalized to "the entire Internet". This is a fallacy.

The Internet has more bandwidth and more points of interconnection than it has ever had before. There are more Internet users, of course, but the average Internet experience today is worlds better than it has ever been. More engineering talent is focused on ensuring the Internet continues to grow than has been involved in any other communications industry in history.

Specific jams
Just like a system of roads and bridges, there are so many places a jam can occur, it is nearly impossible to point out all of them. There can be busy signals at a provider's POP (Point of Presence, the modem pool you dial into to connect to the Internet), not enough network bandwidth at a provider's POP, overload at the telephone company (calls don't even make it to the POP), overload at a NAP (Network Access Point, the place where networks meet to exchange traffic), or congestion and routing failure on a particular backbone.

Quite often, a popular Web site gets overloaded and becomes slow, which can also be equated to a "traffic jam". Proprietary networks such as AOL can also go down. Though this doesn't really have anything to do with the Internet, it is nonetheless lumped in by reporters who don't know any better.

All of these things can affect the Internet experience of a user in one way or another.

Yes, there are problems across the Internet at various times for various users on various Internet networks. But there is no overall "collapse" of the Internet. If there were an automobile traffic jam in Washington D.C. and in Los Angeles at the same time, would it then be accurate to say, "All of America has catastrophic traffic jams"? No. There would be two specific situations, and you would probably laugh at the reporter who tried to generalize these into a nationwide catastrophe.

Unfortunately, due to a lack of understanding about the Internet, and due to journalists' natural tendency toward sensationalism, few people are describing the specific, real problems on the Internet.

What EarthLink is doing about it
Besides having our heads screwed on straight and telling it like it is to anyone who asks, there are many things EarthLink is doing to prevent the kinds of traffic jams mentioned above.

1. EarthLink doesn't rely on one particular national network. We currently supply our customers with dial-in access via our own network here in California and UUNET's network everywhere else. We have just contracted with PSINet to add another 230 POPs to our network. Between these three, we have the largest, most reliable Internet network in existence. This means EarthLink is able to provide alternate POPs when POPs in a particular location get congested. This means fewer busy signals for our customers.

2. All of EarthLink's POPs have plenty of available bandwidth to our networks. This means that EarthLink provides a wide "pipe" from the modems you dial into to EarthLink's Internet networks. Most often, our POPs are connected to the rest of a particular network via T3 (45 Megabit) lines, the largest pipes widely available. EarthLink's customers rarely experience a bottleneck getting to the network from the POP they're dialed into.

3. EarthLink's customers are dialing into private networks. Whether they are ours, UUNET's or PSINet's, these networks are owned and controlled by EarthLink or by EarthLink's partners. Our customers' packets (units of bits, or information being transmitted) travel on these private networks for as long as possible before being handed off to other Internet networks we don't control.

   In contrast, many local ISPs, though their customers dial into their own, local POPs, immediately hand off their customers' packets to public networks run by companies such as MCI and Sprint. Whereas the packets of EarthLink customers could traverse the world without leaving one of our networks.

4. EarthLink has more interconnect points than anyone. Between EarthLink, PSINet and UUNET, we are interconnected at every NAP or other important interconnection point on the Internet. This means that if a customer's packet has to leave our network to get to another network, it will happen at the closest interconnect point providing the best performance and routing around potential jams.

5. EarthLink has a world-class technical staff ensuring our server-based services, such as mail, news, Web and others, grow ahead of demand. We cannot control the response time of servers other than ours, but we can do a good job ensuring our own servers keep up with demand.

   Because of these points, EarthLink is uniquely positioned to keep up with the demand of the growing Internet community and to predict and avoid trouble areas.

   Other Internet service providers have made different choices about how to utilize their resources and are subject to isolated problems much more readily than we are. Netcom's customers, for example, are all dialing into Netcom's single network. These customers' Internet experience is limited to Netcom's ability to grow and manage their network. If Netcom can't keep up, their customers suffer.

   Overall, however, the Internet community and the vast engineering resources behind it are finding solutions to the challenges facing the network. As it continues to grow, the Internet will be stronger and healthier than ever before.

Summary
The Internet's continued high-speed growth creates a challenge to stay ahead of demand. Frequently, this results in isolated traffic jams at particular points on the Internet network. Overall though, the Net is healthier than it's ever been, providing a positive experience for more users than ever.

When talking about problems on the Internet, we must deal in specifics: specific problems, who they affect specifically, and how they can be handled. Generalities born out of misinformation and sensationalism do everyone a disservice. Not only do these generalities fail to describe real issues, they focus attention away from what we are all trying to do: expand the Internet's resources to fulfill the potential of this amazing global communications medium.

EarthLink is positioned to predict and keep up with new Internet demands. Our goal is to give our customers the best possible Internet experience. We are working very hard to ensure this continues.