Though this article is several years old, it provides a practical explanation of why it is still important to build web sites that work well on slow modems. As of 2004, slow, dial-up modems are still used by almost half the households in the U.S.

The Myth of Broadband

John C. Dvorak | PC Magazine | June 12, 2001

Personally, I don't see broadband dominating the online scene anytime soon; an unspoken, de facto standard stands in the way. Everyone talks a big game when it comes to standards, but how many of us understand what real standards are? With computers, we think of Microsoft and Windows right away. When we dig deeper, we might think of USB or maybe IEEE 1394. Bluetooth and other committee-designed or officially designated standards crop up a lot, too.

There are other standards, however, which evolve on their own with no help from any company or governing body. These de facto standards are often ignored by the marketplace and result in failed business plans. Right now, the most important example is the standard Internet bandwidth I/O of around 34 Kbps. Ignoring this number is ignoring reality.

34 Kbps, the typical speed produced by a dial-up connection (plus or minus 10 Kbps), is a true standard. It's easily as important as Microsoft Windows, as many bankrupt companies that banked on broadband can testify. Standards such as Windows are established by the economic model of increasing returns. As the popularity increases, OSs attract more developers because of the larger market size. This scenario doesn't apply only to operating systems. It can apply to processors, applications (Word), or even bandwidth.

The concept that bandwidth itself is a standard is not generally acknowledged. Bandwidth is mistak-enly seen as something that is moving ahead, and if you can jump ahead of the so-called learning curve, you can get a jump on the competition. But bandwidth, like Windows, is a standard, not a product. And also like Windows, it does not follow the learning curve. Windows is not on any learning curve of change. The OS gets upgraded and debugged, yes. But that's all. Standards are maintained, not changed—or at least, they are not changed easily. This is why broadband is going to be a long time coming.

This occurred to me at a recent ISPCon event in Baltimore, where Charles Ardai, CEO of Juno, told me that dial-up subscribers are increasing in numbers, not decreasing. With all this talk of broadband, you'd think that the tide would have turned by now. And when Web site operators or Internet-centric companies look at the facts, they are forced to design and optimize systems for dial-up users. Because dial-up is so cheap and with services designed for it, you end up with the increasing-returns phenomenon. Compare this situation with Windows and Linux: Windows is dial-up, and Linux is broadband—a niche market.

Most activity targets the dial-up user, making broadband just a luxury. And in some cases, it may be an unnecessary luxury, as full-speed feeds to broadband users are fairly rare. I have a megabit line into my home office, and when I view a streaming video feed, I still get a herky-jerky 20-Kbps stream. The true advantage of broadband is realized only on FTP sites or peer-to-peer, where downloading is optimized for speed.

As for surfing, I prefer broadband, because the overall Web experience is improved. But I've noticed recently that the dial-up connection I experience when traveling is not as dreadful as it was in the past. The backbones have improved, and dial-up is more tolerable than before. I think this indicates the standardization phenomenon taking hold—at around 34K.

What does this tell me about the future of broadband? A typical DSL connection costs about $600 a year—something not everyone can afford. We heavy Internet users see things differently and assume that everyone wants to be like us. But the AOL phenomenon should give us pause. Technology mavens saw AOL as training wheels for the Internet, yet AOL now dominates the online world, with over 20 million users—many of whom still use dial-up.

Much of the DSL service is sold to small offices, where it is shared among lots of employees. Sure, there are a few of us with DSL or fixed wireless and a few million homes with shared cable. But it's still a niche (like Linux) when compared with dial-up users. None of this bodes well for broadband becoming commonplace. Nobody today can produce a Web site and not care about the dial-up user. So everything gets designed and optimized for the lowest common denominator: 34K.

It's very difficult to unseat a standard once it is established. Broadband may be decades away.