"I'm someone who really believes in standards, especially Ethernet, and the importance of getting the industry to consolidate around a single technology suite and solve problems in a way that becomes the best cost/performance solution available," he said. Getting to that standard, however, wasn't easy.

Sonet is, by definition, a synchronous system. It gets its timing off the Cesium clock running in Boulder, Colo., and everything that happens occurs in lock step to keep data integrity and make sure all switching occurs in sync and that bits don't run into each other.

"This is an expensive process relative to Ethernet, which buffers and forwards with intelligent packets that don't care when or how, they just go." He said, "Sonet was fine when you only had voice, but as we move to everything being packet, everything being data, it becomes irrelevant."

The solution, according to Thatcher, is to leverage off the installed base of Ethernet networks, which are already permeating the metropolitan area network (MAN) at the latest 1-Gbit/s rates. Then, by developing a 10-Gbit/s Ethernet standard, network managers will be able to extend into the WAN using Ethernet as the end-to-end Layer 2 transport.

While this sounds straightforward enough, implementing it is a different story. When he chose to chair the IEEE P802.3ae, 10-Gigabit Ethernet Task Force when it was formed in March 1999, he had no idea what he was in for. He essentially entered what was to be a monumental clash between the datacom and telecom worlds, as each tried to minimize the compromises they would have to make in reaching an agreeable standard for the physical layer that would allow Ethernet packets to flow smoothly over the Sonet infrastructure. The ensuing debate called upon all of Thatcher's technical and political skills.

"There are a number of reasons why I thought it was appropriate to start up IEEE 802.3ae, or 10G. One was that there were several companies starting to work on proprietary solutions. And once the industry had those established, I thought it would be difficult to move 10G into its proper place," he said.

"OC-192 already existed, and hardware was being installed to a large degree. There were a number of companies that were highly motivated and interested in being able to use the Sonet OC-192 infrastructure to carry Ethernet frames," he said, pointing to work going on in ANSI to do Ethernet over Sonet, which is basically any Ethernet speed over any Sonet speed. "That's not what we' re doing. We're doing something that is 10.000 Gbits/s, and trying to get that married with 9.958464 Mbit/s Sonet, which is really close, so potentially the same hardware could do both."

Because of 10-Gbit Ethernet's close ties to Sonet, many have claimed that 10G isn't Ethernet at all, that the 802.3ae group has simply adopted Sonet as their own. Thatcher flatly denies these claims, "That couldn't be further from the truth," he said. "What we have done is develop a set of PMDs (optical transceivers), the intent of which is to operate in both LAN and WAN mode. The difference is in the encoding, or whether or not the PMD has a Sonet framer or not."

Settling on those PMDs and the encoding schemes was no mean feat, and took up much of the standards development to date. "WAN and LAN networking have never been on good terms and both would just as soon have the other disappear," said Thatcher.

The standard still has another year and a half to go before it is ready for prime time, though a first draft is almost ready for review. In the meantime it is expected that preemptive products will appear before then, in an effort to meet burgeoning user bandwidth demands.

The effect on the average Internet user and the enterprise "will be profound," said Thatcher. "We are seeing, on a more frequent basis, that companies are developing specialty optics to do Gbit Ethernet at 40, 80 or 100 km, so they can install their own private infrastructures and save huge amounts of money. A pure efficiency statement."

While he doesn't predict the demise of OC-192, he sees it as a less dynamic part of the industry. As Ethernet develops, Thatcher sees its scalability as the path to eliminating the Internet backbone bottlenecks.

"Once the price points start to drop for 10G by 2003/2004, there won't be another technology out there with the cost/performance of Ethernet. It just won' t be there," he said.