That is a seemingly odd position for the Stanford professor and Amati Communications cofounder, who is credited with the invention of the asymmetrical digital subscriber line. ADSL, as pioneered by Stanford and Amati (now part of Texas Instruments' Internet Access Group), uses discrete multitone (DMT) signaling that effectively divides the POTS telephone line into 256 parallel adaptive segments (like parallel modems), each capable of a 32-kbit/second symbol rate. Thus, ADSL uses standard phone lines to transmit data at up to 8 Mbits/s over a 16,000- or 18,000-foot range.

When Amati was formed in the early mid-1990s, there was a great deal of controversy over the appropriate ratio between asymmetrical service types, Cioffi said. The decision to go asymmetrical was based on the supposition that upstream data transmissions-especially if they consisted primarily of typed commands from a keyboard-would seem puny next to the massive data dumps (a digitized two-hour movie, for example) that Internet service providers anticipated sending downstream.

The AT&T group that became Lucent Technologies advocated a 100:1 download-to-upload ratio, while GTE believed in something closer to 1:1, Cioffi explained. Amati was going for an 8:1 ratio; G.Lite (the splitterless version of ADSL) uses a 10:1 ratio (1.5 Mbits/s coming down from the host and 150 kbits/s going up).

Semiconductor companies, such as Lucent and Broadcom, and service providers, like Qwest, still see movie downloads as the primary reason for utilizing asymmetrical subscriber lines, Cioffi said. But Internet service providers are banking on fiber optic cables getting closer to the home (at least to the curb), though the last hundred feet or more would still be standard twisted pairs. And telephone companies are increasingly seeing more potential in symmetrical services. "They'll charge for it," Cioffi said.

Cioffi is a member of the National Research Council Computer Science and Telecommunications Board, and serves on the technical advisory boards of a number of companies including Marvell, Coppercom, Gigabit Wireless and Charter Venture Capital. As chairperson of the VDSL Alliance he heads a group that includes telecommunications companies like Alcatel and Ericsson, network companies like Cisco Systems, and semiconductor manufacturers like Mitel, NEC, STMicroelectronics and Texas Instruments.

VDSL (Very high-rate Digital Subscriber Line) is a new specification that will soon be an American standard, said Cioffi. The data transmission will be close to 26 Mbits/s symmetrical (25.92 to 27.6 Mbits/s), or 50 Mbits/s (51.84 to 55.2 Mbits/s) asymmetrical "on a short line" (up to 1,000 feet), said Cioffi. As with ADSL, the data rate will depend on the distance from the central office, trailing off as the data receiver gets further away. Thus, the downstream rate is approximately 13 Mbits/s (12.96 to 13.8 Mbits/s) at 4,500 feet from the central office.

VDSL promotes itself as easier to implement, and much higher speed-but with a smaller distance from the central office. Development efforts on VDSL now concentrate on physical-layer implementation, Cioffi said. VDSL send and receive carriers would be out of the range of voice or ISDN services, thus allowing phone companies to overlap data services on the same line. VDSL, which also depends on frequency-division multiplexing, is intended to be backwards compatible with ADSL, said Cioffi.

Cioffi can trace progress from 1987 when Joe Lechleider, a research director at Bellcore, encouraged Cioffi and his Stanford students to "go for it," he said."We were the only two guys who believed in DSL," said Cioffi. Bellcore funded Stanford's DSL research in the early '90s, and Alcatel was the first to license the technology, he said.

The full-rate (8-Mbits/s) early Amati modems used opto-isolators and cost $10,000 each. They required splitters (bandpass filters) to insure the analog voice and DMT data lines were separate. Though increasingly voice is transmitted digitally, the "trick" the modems need to master was a complete reversion to POTS technology in the event of a power failure. That is, the voice would become analog again if the DSL system went down. And data rates would deteriorate with noise on the line-a likely occurrence over a four- or five-mile span.

Even now, like the final rollout of ISDN, getting DSL services can still be a problem-even in California's Silicon Valley. "I'm the inventor of ADSL, but I had to call Pac Bell four times just to get it," he said.

Cioffi admits there is criticism of the telecom industry for the slowness with which it has deployed ISDN and now DSL services. "The computer industry will say the phone companies are slow-and there is pressure from alternate services," he said. "But this is a massive new service being deployed."

But fixed wireless (using 2- to 5-GHz carriers for broadband access) is far behind cable and DSL service development and deployment, Cioffi said. Cable modems, similarly, have proved themselves a good vehicle for TV distribution, but perhaps not for Internet access. The problem is that connection is not point-to-point; it is shared, which could lead to problems with upstream bandwidth. To be sure, the National Research Council in which Cioffi participates sees a convergence in the way services come to the home, and that is through fiber optic cable. "It will all migrate to the same solution in the long run," he said.

Not surprising, Cioffi still sees the telecom infrastructure-and DSL services-as offering the best vehicles for expanded broadband access. "Clearly fiber is the long-term solution-it will be everywhere-but that could take 100 years." In the meantime, twisted-pair phone lines remain the vehicle of choice for "squeezing out the last few bits," he said.