The group — which recently held its third meeting, in Hilton Head, S.C. — has approved a project authorization request and several areas of focus that by the fall could be approved by other committees involved in various aspects of the IEEE 802.3 (Ethernet) standard. Once that approval is received, the proposed Ethernet in the First Mile (EFM) Study Group would become a task group to hammer out the specification and offer it for public comment and voting. The process could be completed by early 2003, but early products are likely to bubble up before that.

"There's a very, very strong feeling that the progress we've made is sufficient" for approval from the other committees, said Brian Murray, chief technology officer of Massana Inc., who attended the meeting. "Most people in the group realize we need to standardize all those elements to get one cohesive standard together . . . for a winning solution."

With Ethernet technology having taken over 90 percent of LANs and now finding use in the network core at gigabit rates, the natural next step is to bring it all the way into the home, with the high bandwidth promising voice, video and data over existing copper. Residential and small-business access to networks is predicted to grow to 40 million nodes in the United States by 2005, according to data supplied by Dominet Inc.'s Howard Frazier, chairman of the study group.

"Ethernet consistently demonstrates the most attractive cost/performance ratio of any networking technology, at any operating speed," Frazier said.

In addition, the TCP/IP stack that drives today's Internet fits naturally on top of Ethernet, said Patrick Stanley of Elastic Networks.

The devil will be in the details, but the first debates on a few of them appeared to go smoothly. The 130 attendees at the March 13 meeting overwhelmingly approved four focus points to send to the other 802.3 committees, Murray said.

In general, the group is looking at a hybrid approach to bringing Ethernet to the home (what some call the last mile, the group now calls the first mile). The members approved, with little debate, a focus on long-distance fiber physical-access layers that would allow point-to-point single-mode fiber or point-to-multipoint fiber in distances of 10 to 20 km.

John George of Lucent Technologies argued in a presentation that such an approach allows data rates of 1 Gbit/second from the central office to a local multiplexer and rates of 100 Mbits/s and above from the multiplexer to the user site. Murray acknowledged that neighborhoods may initially have one fiber serving 500 customers at 10-Mbit rates. But as uptake rates increase and fiber gets closer to the home, 100 Mbits will be attained, he said.

And while fiber-to-the-home installations are forecast to grow to 20 million by the end of the decade, the group approved a focus on a copper physical-access layer in a nod to the enormous installed base around the world (one forecast estimates that only 5 to 7 percent of the market today is served by fiber to the home). There will be crosstalk issues, but Murray said algorithms that address them will be worked out when the nitty-gritty spec work gets under way.

"There's so much copper in the local-loop and multiple-dwelling markets that you're going to see copper connect to the end user," he said.

The group approved a third focus on a physical-access layer for Ethernet over passive optical networks and a fourth on operational administration maintenance and provisioning. The last acknowledges the brave new world that Ethernet would encounter in the home.

"It's not a LAN in a building," Murray said. "When you get into a subscriber network, the people who are using it are customers. It's a more hostile environment. They need a much, much greater amount of support."

Supporters argue that the combination of copper and fiber and an ease-of-use model should help displace digital subscriber lines and cable modems over time. DSL users need to be located no more than 2 miles from a telecommunications cabinet or pedestal, and cable modems often have slow upstream data rates.

Entrenched competition

On the other hand, those broadband technologies are available today. DSL rates, depending on whether the approach is symmetrical or asymmetrical, can be as slow as a few hundred kilobits per second (the floor of what the FCC defines as "broadband") to up to 8 Mbits/s on the downstream. Downstream speeds on cable modems can range theoretically from 3 to 30 Mbits/s but practically range from about 400 to 1,400 kbits/s.

DSL buildout in the United States reached 2.4 million homes in the fourth quarter and is growing by at least 40 percent per quarter, according to market research firm Telechoice. DSL's reach is forecast to extend into more than 17 million homes by 2004. And some market watchers say as many as 4 million U.S. homes have cable modems today.

Any Ethernet-to-the-home specification must demonstrate significant advantages over those existing technologies, skeptics argue.

Murray acknowledged that "there's a market for [DSL] today," but maintained that "it's not something that will get us the data rates we need over the next 10 to 15 years."

He described the experience of a local-exchange carrier from Hilton Head that had built out a DSL-type network in the area. A representative of the carrier had spoken at the meeting, Murray said, and reported that "every connection had required a technician and tweaking. After two to three years, they pulled out [the DSL network] and put in an Ethernet packet network. Suddenly they were able to do all provisioning over the data channels."

If the approach seems long overdue, only recent advances in signal processing and CMOS technology have made deploying Ethernet to the home practical. But the architecture leverages what's proven and in high volume: Ethernet silicon. (The standard emerged from the Xerox Palo Alto Research Center in the '70s using carrier-sense multiple-access collision detect to run over coaxial cable.) Products could emerge late next year that include gateway chips for customer boxes and multiport switches, Murray said.

"It's no longer a question of whether, when or where" Ethernet comes to the home, Michael Silverton of Fiberhood Networks (Palo Alto, Calif.) said in a presentation to the group. "It's merely a question of where [Ethernet to the home] is building next and how to ensure that standards are enforced and identities preserved."

From the service providers' standpoint, Ethernet to the home should open a land-office business in "cafeteria services," according to Brad Booth of Intel Corp., the group's chief editor.

An Ethernet-to-the-home network architecture creates a point of access for users that allows easy, one-step access to providers of voice, video, Internet services and other applications, he said in his presentation. That business model would be a boon to service providers that aren't making much money on fixed-price DSL services.

The flexibility of such an architecture is enormous. Over one connection, a user could conceivably run an Internet hookup at 6 Mbits/s, four concurrent telephone calls at 0.064 Mbit/s each and four concurrent videophone conversations at up to 768 kbits/s each, according to Richard C. Brand, director of network architecture and applications for Nortel Networks.