Network Computingwww.networkcomputing.com

The deployment of xDSL is not the next chapter in the glorious history of the modem business. It's the first chapter in the brave new world of public network computing infrastructure--the rules of which are still being written.

Where It All Began The idea of squeezing more usable bandwidth out of the billions of dollars of copper buried in the ground has a long heritage, predating the very idea of a PC. The original concept was simply one of pair gain--that is, providing two or more plain old telephone service (POTS) connections over a single line.

The first widespread embodiment of this concept was Integrated Services Digital Network, or ISDN, the neglected stepchild of post-Internet digerati. The parameters for ISDN were chosen such that two toll-quality voice conversations could be carried over most local loops, with a smidgen of bandwidth left ove r for supervisory signaling. Could higher data rates have been chosen? Sure, but a conservative approach allowed phone companies to develop predictable deployment plans for uniform tariffed services without driving their operational people bonkers dealing with cu stomers on loops that couldn't work.

When the PC industry came along in the early 1980s and decided it would be cool to transmit data over the phone network, did it rally around the phone company trying to figure out how to best use ISDN for data? Of course not. At the pace the phone company was deploying ISDN, this would have been suicide. Instead, voice modems were invented that operated within the end-to-end bandwidth constraints of the Public Switched Telephone Network (PSTN), treating it like a transparent cloud. This allowed the computer industry to establish widespread retail distribution without giving a thought to what the phone company was or was not doing. What followed was an unimpeded march from 2.4 Kbps to 33.8 Kbps, with a final su rge to 56 Kbps--at least downstream if the service provider is on a digital line, and all the planets are in alignment.

So here we are in the closing years of the '90s, in a world filled with top-notch digital signal processing wizards who know they can do better, egged on by a trade press that still treats the PSTN like a transparent cloud. Unfortunately, most of these folks have probably never seen the inside of a telephone central office, much less climbed down a manhole to take a gander at what copper looks like in the wild. Sure enough, bright engineers can build gizmos that spit multimegabit data streams down a twisted-pair running between the lab and the conference room.

But lo and behold, take these devices out into the real world, and your mileage may vary. By pushing the limits of physics, xDSL--at least the higher-speed versions--comes with fairly onerous limitations on loop lengths and line conditions. These limitations easily can render a majority of the installed base of copper pairs unu sable. A small problem of implementation? Well, not exactly.

Watch Those Speed Bumps Here's a little thought experiment for you. Imagine you're getting ready to introduce a new processor chip--the Stupendium--that promises to boost PC clock speeds up to 5, 000 MIPS. There's just one minor catch. The performance of the chip depends on the physical characteristics of the PC user's desk. On lab benches, the chip works as advertised. On mahogany credenzas, the chip runs at about 1,000 MIPS. On metal desks, you get maybe 300 MIPS out of it, but only if the desk faces north. And if you put the PC on a folding table, it doesn't work at all.