The Original Internet Bottleneck

Many early ISPs—particularly AOL—weren’t ready to offer unlimited internet access in the mid-’90s. That is, until a surprising disruptor appeared: AT&T.

By Ernie Smith

Today in Tedium: Recently, there’s been a lot of talk about bottlenecks in relation to the internet. Which makes sense, obviously, because everyone is home trying to access Zoom on a 12-year-old cable modem. Thus far, despite repeated concerns from officials and the public, the internet has largely held up amid the beast that is COVID-19. The real problem, though, is access. Rural environments notably have crappy internet access, dealing with low-speed DSL or satellite access because of regulatory failures that didn’t fill the access gap in time. But going further back, I wanted to spend a little time tonight discussing a historic period where the internet actually buckled a bit—at the provider end. Today’s Tedium is about the bottlenecks the internet faced when dial-up access first went mainstream. Keep dialing, eventually you might get through. — Ernie @ Tedium

Tedium on Patreon

Keep Us Moving! Tedium takes a lot of time to work on and snark wise about. If you want to help us out, we have a Patreon page where you can donate. Keep the issues coming!

We accept advertising, too! Check out this page to learn more.

Let’s talk about this commercial: The one dude comes to the other dude’s house to see if he’s ready to go to the game, only to reveal that he can’t go to the game himself. Why did you go to his house? This commerical is built on a logical fallacy.

The day AOL opened up the floodgates to the internet

Real internet users have long had deep suspicion of America Online, because of what it represented. It wasn’t the “real” internet—rather than forcing you to use something like Trumpet Winsock to manage your connection, or relying on your terminal skills to log in, you were given a friendly interface, and they called the shots. Given the technically savvy culture that berthed the internet, it was an easy target.

Decades later, our major social networks would look a lot like AOL, but our ISPs would not. And a lot of the reason for that comes down to a pivotal decision AOL made on December 1, 1996. That was the day it offered flat-fee unlimited access to its service for the first time.

Previously, the company offered different tiers, with the most popular being 20 hours a month, and a cost of $3 for every extra hour.

Announced the month prior, AOL both revealed that people could stay online for as long as they wanted after paying a flat $19.99 per month, but also that they that it would upgrade its access technology so its users could use a real web browser, rather than the built-in web browser that the service offered users. As Chicago Tribune columnist James Coates noted at the time, the change also added Windows 95 support to the service, which let the company “transform itself into a full-blown 32-bit Internet service provider at a flat, untimed $20-per-month toll.” (The horror of having to use software designed for Windows 3.1 to surf the web in Windows 95! The horror!)

But this move was a pendulum that swung both ways. For months after AOL made its move, trying to get access to the company’s network was near impossible—busy lines were a constant presence. Some people would try to get around the issue by buying a separate phone line and keeping the line as occupied as possible, so they wouldn’t have to dial back in. Dialing back in was torture. There was a vast digital sea to access—if you could get the initial handshake to happen.

0409 aoldisks

It probably did not help that AOL distributed so many disks to users throughout the mid-1990s. (monkerino/Flickr)

Less appreciated at the time was how significant a change this was to AOL’s business model. In one fell swoop, the largest online service provider in the world was opening up to the full internet and moving its business model away from the carrot-and-stick setup of most online services up to that time.

Before that moment, online services like AOL, along with its earlier forebears like CompuServe and Prodigy, had pricing models that were based on how much you used, and they actually became less expensive with time, not more. The companies, notably, inherited a pricing strategy used by early bulletin boards and digital access platforms such as Dow Jones’ online information service, which charged for content by the hour on top of a monthly fee. This was not exactly a friendly model for consumers, and put a barrier up for the kind of addictive access we rely on today.

There were other bottlenecks as well, of course. Modems were slow on both ends of the equation—2400 and 9600-baud modems were common in the mid-1990s, and speeds were artificially limited by the quality of the connections on the other side of the line. You could have a 28.8 kilobit modem, but if the best your online provider could do was 9600 baud, you were kind of stuck.

Perhaps the biggest bottleneck to always-on access, however, was a business one. Early internet providers simply did not know if it made sense to give us more internet access—or if the business model would make sense if they weren’t charging us hourly for the right to get online. And they had infrastructure concerns, too: If you offered everyone unlimited internet, you better have enough infrastructure to handle all of those calls.

In the 2016 book How the Internet Became Commercial: Innovation, Privatization, and the Birth of a New Network, author Shane Greenstein explains how the pricing of internet access was a major challenge, and it was not immediately clear who would have the winning argument for the internet era. Here’s how Greenstein explained the two philosophical camps in ISP-land:

Two broad viewpoints emerged. One viewpoint paid close attention to user complaints about monitoring. Users remarked that surfing the World Wide Web was hypnotic. Users found it challenging to monitor time online. In addition, monitoring time online was nearly impossible with multiple users within one household. The ISPs with sympathy for these user complaints argued that unlimited usage for a fixed monthly price would be a feasible solution. A premium would cover the extra costs of the unlimited usage, though it was an open question how large that premium needed to be. These plans were commonly referred to as flat rate or unlimited plans.

An opposite viewpoint contrasted with the first one. Specifically, user complaints were transitory, and new users had to be “trained” to monitor their own time online. Supporters of this view pointed to cellular telephones and bulletin boards as examples. Cellular telephony was also beginning to grow, and pricing per minute had not deterred users. One up-and-coming bulletin board firm, AOL, had seemed to grow with such usage pricing as well. The ISPs that held to this view expressed confidence that usage prices would prevail, and they called for exploration of new combinations that might better suit surfing behavior by nontechnical users.

This led to something of an awkward state of affairs, where it was not totally clear how much of the golden goose the gatekeepers would be willing to give away. The disruptor that changed everything, somewhat ironically, was AT&T.

0409 worldnet

An early ad from AT&T WorldNet, the first national internet provider that offered flat-fee unlimited access. (via

How AT&T turned unlimited access into a de facto requirement for mainstream internet service

If you know anything about the history of AT&T, they are generally not the ones who break barriers.

Rather, they’re the ones who try to keep the status quo. One just has to look at the history of the TTY system, in which Deaf hackers that wanted a way to communicate with their friends essentially invented the acoustic coupler (a gadget in which you literally put a phone on top of a microphone and speaker) to get around the fact that Ma Bell didn’t allow third-party devices to plug into their phone lines.

But a lot had changed by early 1996, when AT&T launched its WorldNet service. The RJ11 telephone jack, which nearly all modems used by the early 1990s, came about as a result of a legal decision that blocked AT&T from limiting the use of third-party peripherals, something that gave us answering machines, cordless phones, and … modems.

But by the beginning of 1996, the company found itself in its odd position as a disruptor of the then-still-young internet industry. It was big enough that people who had never used an ISP might jump on board, and could attract heavy users by charging a flat fee—$19.95 for unlimited access if you had long-distance service with the company, $24.95 if you didn’t. To sweeten the deal, the company offered users five free hours of internet access per month during its first year. (And notably, it offered 28.8 kilobit speeds, relatively fast for the time.)

The problem, per Greenstein, was that it was a bet on scale. By charging so low for internet access, they were effectively hoping that tens of millions of people would sign up for WorldNet—and if they couldn’t guarantee that, it wasn’t going to work. “AT&T took a calculated business risk by choosing a design for its service that would not—indeed, could not—make a profit if it did not get widely adopted in many cities throughout the United States,” he wrote.

AT&T was not the first to flat-fee access—personally, I used an internet provider that offered unlimited dial-up access as far back as 1994, a reaction to my extreme enthusiasm for long-distance bulletin boards that eventually showed itself on my parents’ phone bill—but AT&T was so big that it could handle a national launch of flat-fee internet service provider in a way that a smaller regional competitor might not.

A New York Times article by famed tech writer John Markoff noted that AT&T at one point had its sights on building an AOL-style walled garden of its own, much like AOL or what Microsoft did with MSN. But sometime in 1995, it decided instead to just give people a pipe to the internet, using open standards.

Markoff wrote: “If AT&T builds an attractive and inexpensive portal to the Internet, will its customers follow? And if they do, will anything in the communications industry be the same?”

The answer to that second question, of course, was no. But not necessarily because of AT&T, though it did get a whole lot of customers thanks to its decision to price internet access in an unlimited fashion. Really, it was the reaction to the AT&T’s entry into the market, which set a new baseline for internet access, that changed the industry forever.

Expectations were reset. Now, every ISP in the country was in a position where it needed to offer unlimited access and match WorldNet’s price, just to keep up.

As Greenstein notes in his book, this had a disruptive effect on the still-young ISP industry, with AOL and MSN the only services large enough to hold out on the price front. (Notably, CompuServe blinked, launching its Sprynet service at the same flat-rate $19.95 price as WorldNet.) AT&T’s move also angered the Baby Bells, because of a decade-old FCC decision that allowed the company’s data lines to get around access charge rules it was otherwise required to pay for local voice calls.

AOL, with a significant content business that lived in a proprietary sphere, initially tried to play both sides, offering a discounted version of its service that lived on top of AT&T’s connection.

But soon, it too would have to fold to the new normal—the flat-fee requirement of dial-up internet access. But that decision brought a whole lot of pain.


The call failure rate for AOL, according to a spring 1997 study by Inverse, a web measurement firm. The number was roughly double the next-worst offender, and likely a result of a poorly optimized network of dial-in banks. In comparison, CompuServe (which was the most successful company in the study) had failure rates of 6.5 percent.

0409 modem

A 28.8-kilobit modem, which was highly coveted among home internet users for a time in the mid-1990s. (Les Orchard/Flickr)

Taming the busy signals: Why getting online was such a nightmare in 1997

The question we’ve been hearing frequently over the last few weeks—can the internet keep up with our increased use?—was the same question that was being asked in early 1997, as more and more people flooded online for hours on end.

And the answer appeared to be no—not necessarily because websites were hard to access due to the extra interest, but because phone lines were.

(Infamously, individual websites got their big-news stress test on September 11, 2001, as the internet buckled under the stress of a flood of interest in a major news story, as well as the destruction of significant amounts of infrastructure in one of the largest cities in the world.)

And AOL’s infrastructure, already stressed due to the popularity of its service, was simply not designed for the extra haul. In January of 1997, less than months after going all-in on unlimited access, the company found itself, under pressure of attorneys general around the country, having to promise refunds and limit its coaster-friendly advertising until it could fix its infrastructure problem.

AOL roughly doubled the number of modems available to subscribers, noted The Baltimore Sun, but it was clear there was something a little bigger at play, obvious to anyone who ever used the phone system to dial into a data-based service and got a busy signal: The phone system wasn’t designed for this, and it was becoming painfully obvious.

The Sun article noted that the design of the phone system wasn’t built for the lines to be used up 24/7 like dial-up modems encouraged by their design, and this stress on the phone system was pushing the Baby Bells (unsuccessfully) to make the case for additional usage fees. The FCC wasn’t having it, which meant that the only real solution for this congestion was for a replacement technology to come along to usurp those phone lines—which was what eventually happened.

“We use the regular telephone network because it’s there,” author Michael J. Himowitz wrote. “For delivering data, it’s slow and inefficient, and there isn’t necessarily any reason that the needs of Internet users should conflict with those of voice callers.”

That meant that, for a few years at least, we were stuck with a completely unsustainable system, one that negatively affected everyone, not just AOL users. While it’s not clear whether he was an AOL user or reliant on another service, Todd Rundgren somewhat infamously wrote a song about the anger and frustration one would feel when unable to connect to their internet provider, a tune called “I Hate My Frickin’ ISP.”

ISPs tried baking in alternative business models to convince users to use the internet less, Greenstein wrote, trying to charge less or pushing a particularly aggressive user to use another service by refusing them unlimited access. But once the cat was out of the bag, it was clear unlimited access was the way things would be.

“Once the market moved to that general practice, ISPs could not find many takers for the alternatives,” wrote Greenstein. “Competitive forces kept matters oriented toward the user’s preferences for unlimited capacity.”

AT&T WorldNet was not immune to the problems unlimited internet service caused. By March of 1998, just two years after its service launched, it announced that it would start charging users 99 cents an hour for each hour they used above 150 each month. 150 hours is still decently reasonable—roughly equivalent of five hours per day, what might happen if you spend your entire evening on the internet every night instead of watching Friends—but definitely a backing down from the promises of “unlimited” internet.

As for AOL, they might have come up with the best possible solution to this awkward competitive situation: After spending hundreds of millions of dollars upgrading its infrastructure, the company bought CompuServe in 1997, effectively doubling its dial-up service overnight. Around the same time, the company sold its dial-up facilities and contracted them out, according to Greenstein, ensuring that the busy signals would be someone else’s problem.

Which is pretty genius, if you think about it.

It seems obvious now that it was destined that we were somehow going to get unlimited internet access.

After all, I have to imagine college students using T1 lines in the dorms being extremely disappointed in the off-campus alternatives. The disparity was so strong that there was no way it was going to hold forever. We would expect unfettered access to this series of tubes if we were to be productive members of society.

(Mark my words: There were probably a sizable number of people who went to college in the ’90s and early 2000s that extended their stay at their school of choice just because they wanted access to the then-rare high-speed internet available at those schools. Double-major? Sure, if it’ll keep my download speeds snappy!)

The experience in the dorms might have been amazing, but dial-up modems were quite obviously not up to the task of bringing that experience home. But the failings of the dial-up system eventually led to the growth of better technologies, with DSL (which used existing telephone lines to transmit high-speed data) and cable-based internet (using common lines that themselves took time to be added to homes) helping to get most users closer to the high-speed internet options once exclusive to college campuses.

As I was writing this, I found myself thinking about what the world might have been like had a disease like COVID-19 hit when we were trying to get online largely with dial-up, given that diseases like it seem to appear once every century. Would it have been easy to go remote like it’s proving now? Would the busy signals have prevented the economy from continuing to move? If AOL hid dial-in numbers from its users like they were accused of, would it have led to rioting?

Would we even be able to get delivery like we are now?

I don’t know the answers to these questions, but I do know that, as far as the internet is concerned, if we were going to be forced to stay in place like this, from a communications standpoint, now is basically the best possible time for something like this to happen.

I could not imagine the added stress a busy signal would add on top of all of the other quarantine concerns right now.


Be sure to check out the book I featured in this issue, Shane Greenstein’s How the Internet Became Commercial: Innovation, Privatization, and the Birth of a New Network. It’s quite good.

Find this one an interesting read? Share it with a pal!

Ernie Smith

Your time was just wasted by Ernie Smith

Ernie Smith is the editor of Tedium, and an active internet snarker. Between his many internet side projects, he finds time to hang out with his wife Cat, who's funnier than he is.

Find me on: Website Twitter