My first modem was built into a Teletype ASR33 terminal attached to the PDP-8 minicomputer that heated my house in Palo Alto for many years. That modem ran at 110 baud or 10 characters per second. My first microcomputer modem was a $300 Hayes Smartmodem rated at 300 baud, which seemed plenty for the primitive bulletin board systems of the late 1970s. Today I'm sucking data down a T1 line with more than 5000 times the capacity of that Hayes modem and find myself complaining about a lack of bandwidth. What has changed, of course, is what I do with the bandwidth. Instead of ASCII text files, I'm downloading Linux distributions and streaming video just like everyone else. And just like me, everyone else is running out of bandwidth -- a condition that the computer industry is doing little to remedy.

Given that there are something like 60 million U.S. homes on the Internet and fewer than 20 million of those have broadband connections, you might think I am a bit premature in my complaining about a lack of bandwidth. After all, compared with two-thirds of Americans on the Net, I have plenty of bandwidth, tons of it. But my complaint isn't so much with the amount of bandwidth any of us have, but with the amount that is available. I fear we are falling behind the bandwidth curve and this is going to severely limit the potential of the Internet.

When Bob Metcalfe and his co-workers at Xerox PARC were inventing Ethernet in the 1970s, their major concern was how to dump 600 dots-per-inch bitmaps from a minicomputer to a laser printer. The printer could run at several pages per minute, but the ancient parallel port connecting the printer to the computer required 15 MINUTES to send each page. Something better was needed, and that something was Ethernet. Able to transmit data at 2.67 million bits per second across a coaxial cable, Ethernet was a technical triumph, cutting the time to transmit that 600 dpi page from 15 minutes down to 12 seconds.

At 2.67 megabits-per-second (mbps), that first version of Ethernet was a hell of a product, both for connecting computers to printers and, as it turned out, connecting computers to other computers. Every Xerox Alto computer came with Ethernet capability, which meant that each computer had an individual address or name on the network. Each user named his own Alto. John Ellenby, who was in charge of building the Altos, named his machine Gzunda "because it gzunda the desk."

The 2.67 mbps Ethernet technology was robust and relatively simple. But since PARC wasn't supposed to be interested in doing products at all and was devoted instead to expanding the technical envelope, the decision was made to scale Ethernet up to 10 mbps. Metcalfe had done some calculations that suggested the marketplace would need only one mbps through 1990 and 10 mbps through the year 2000, so it was decided to aim straight for the millennium and ignore the fact that 2.67 mbps Ethernet would, by these calculations, have a useful product life span of approximately 20 years.

What we have going here are parallel lines of development. Bob Metcalfe was aiming for 10 mbps at the same point in the late 1970s when I was using that Hayes 300 bps modem. Metcalfe's work was aimed at businesses while most Hayes customers were at home. But a curious thing has happened in recent years as the development of business networking and home networking have intersected. Suddenly the bandwidth requirements of our homes tend to be greater than that of our businesses. Just as gamers now have the fastest PCs, so too advanced home users are shoving far more bits over their home networks and the Internet than do most businesses. The result of this trend is that 10 years ago, the standard for business Internet connectivity was a T1 line at 1.544 mbps, and the standard for home Internet connectivity was a V.34 modem at 28.8 kbps. But 10 years later, the standard for business Internet connectivity is still a T1, and the home standard is now a 384 kbps broadband connection or better.

The problem with these numbers is that they don't fit what's actually happening in the industry, which is now concentrating on building out cable modem and DSL services to reach everyone. This ignores the fact that home network users are going to continue to accelerate, and the development programs to come up with those faster networking products simply don't exist. THERE IS NO SUPER DSL STANDARD IN THE WORKS. Sure, we have our PCs coming equipped with 100 mbps Ethernet and maybe 54 mbps WiFi-g, but we don't connect to the Internet any faster as a result. Look at all the DSL charts and you'll see they tail off at 1.5 mbps, if that. Cable modem specs are more confusing but even drearier, since hundreds or thousands of users often have to share the same 10 megabit-per-second backbone.

The answer that the telephone and cable industries come up with, if they come up with anything at all, is that we'll gain faster speeds through infrastructure changes. We'll have optical fiber to the home, or perhaps hybrid fiber-coax. That's the story they've been telling us for the last decade, but the simple fact is that there isn't that much fiber out there connecting homes. Sure, the odd housing development or apartment complex comes pre-wired, but the numbers are dismal, around 100,000 residential units per year out of the 100+ million total residences in the U.S. At this rate, it will take a century or more for faster connections to get everywhere they are needed. And though I have been taking my vitamins, I am not at all convinced I can wait around to see that happen.

We can't wait for phone companies to replace their copper wires with optical fibers. Going further, I'd say we WON'T wait, because market vacuums tend to be filled, though not always by the people you expect.

This is yet another chance for the telephone industry to screw up. It is their game to lose. For 100+ years, they have been supremely confident that they owned the last mile to our homes and we are stuck with what they offer. Except we aren't stuck. Cable modems have made great inroads, though I don't think they are the next step, either. Nor is wireless networking, which I see abused in all the wrong ways, sometimes because of things I have written. You simply can't put 100 million houses on WiFi links and expect no technical problems.

No, the answer lies in wires and the wires belong to the phone company. This is the case for two reasons. First, nobody is going to pay to replace those wires with fiber. It just costs too darned much money, so fiber goes into new construction, not old. Second, wires operate in parallel so that each connection has all the available bandwidth to itself. My town has 4000 DSL users and 10,000 cable modem users, but which group has greater available bandwidth? The 4000 DSL users average 500 kbps each while the 10,000 cable modem users operate on eight 10 mbps subnetworks. That means there's five gigabits-per-second of available DSL bandwidth, and only 80 megabits-per-second of cable bandwidth. Understand that we are not talking about Internet bandwidth here, but about the local network.

If we are going to get to HDTV video-on-demand in every home, to wall-sized video conferences with Grandma, to the communication utopia described by every so-called expert, it is going to require massive increases in bandwidth. And for reasons of cost, that increase will almost certainly have to come, for the most part, over the existing copper wires in our homes. If we leave it to the phone companies, though, it will never happen.

Fortunately, there is at least one solution to this problem, which I will share with you next week.