Broadband horizon

MOST INTERNET CONNECTIONS that claim to be broadband are really pretty lame. For example, ISDN, which is usually considered the low end of broadband, is barely twice the speed of an analog modem. DSL is just eight to 10 times faster than an analog modem for most users. Even the heavy lifters such as T1 lines are only one-eighth as fast as plain old 10Base-T, an Ethernet standard for LANs (local-area networks) that many companies are abandoning because it's too slow. So how can you say that ISDN, DSL, or even T1 is broadband?

Most connections to the Internet top out at about 1.5Mbps but have been called a "broadband" connection because that's the best most companies could get. Until connections become fast enough to do things fundamentally different from just surfing the Web -- receiving high-definition television, for example -- we aren't really at broadband level. Fortunately, that may change in the near future.

An organization called the Ethernet in the First Mile Alliance (EFMA) is working on a new Ethernet standard designed to use existing infrastructure to provide true broadband access to the Internet. The idea is that if phone companies can use wiring they already have to provide Ethernet at 10Mbps to customers instead of DSL or T1 lines, and do it at significantly lower costs, they will.

To accomplish this, the EFMA is working on a proposed standard, designated 802.3ah by the IEEE. This standard would support Category 3 cable (standard, voice-grade phone wires) to 750 meters. In addition, the standard would support 100Mbps and 1Gbps Ethernet over optical fiber. To keep costs down, the EFMA is also proposing a standard for Ethernet over a single strand of fiber, in addition to the dual-strand fibers currently in use.

Single-strand fiber would work by using different wavelengths of light on the same strand for the two directions, instead of using the same wavelength over two separate strands of fiber, as in the current dual-strand practice. In either case, the EFMA's proposed standard calls for fiber runs as long as 10 kilometers. The current thinking of the 802.3ah committee is that these fibers would aggregate many 10Mb copper drops -- with only half as much fiber.

There's also support for passive optical networks in the 802.3ah proposal. This standard would allow several users of Gigabit Ethernet to share a single fiber for as many as 20 kilometers, using passive splitters to aggregate and separate traffic.

In addition to the normal Ethernet traffic, the EFMA proposal contains a provision for OAM (operation, administration, and maintenance) protocols that the telcos need for any traffic carried over their networks.

Looks fairly straightforward, right? In fact, it is. EFMA members expect a final draft of the standard by the end of the year. Final adoption could happen as early as 2003.

The reason things are proceeding so smoothly is that this is just good ol' Ethernet with some minor changes to accommodate slightly different transmission media and the OAM protocols.

And the differences are slight indeed; after all, 10Base-T has used Category 3 cable for years, and Fast Ethernet and Gigabit Ethernet have been using fiber for a while.

At this point, the only real topic up for discussion is a proposal for long-range copper-based support that could extend as far as 80,000 feet. Whether this proposal becomes part of the proposed standard remains to be seen as the technical details get hashed out. However, the EFMA's other fiber proposals, combined with copper-based drops to customers, are expected to work well with existing telco wiring plants.

In addition to leveraging existing physical assets, the EFMA proposal also provides for a simpler solution than today's DSL. For example, 802.3ah would eliminate the need for the current ATM traffic that telcos use to support Internet access via DSL. This, in turn, reduces the complexity of ATM while increasing efficiency by eliminating frame mismatches between Ethernet and ATM.

The EFMA's 802.3ah proposal seems to have a lot of promise and early support. A few vendors are reportedly already working on products that will leverage 802.3ah, well in advance of the formal standard.

Just how far the telcos plan to go to support the effort is still uncertain. If the EFMA's solution is indeed faster, simpler, and cheaper than existing DSL, then it's hard to see why getting telco support would be an issue -- after all, customers would snap it up.

But remember, a T1 line can cost far more than $1,000 per month, and 802.3ah is many times faster -- and telcos are worried that the new, speedier technology will disrupt their pricing and market structure. At this point, it's not clear whether phone companies have any real incentive to offer lots of bandwidth for little money. Changing that attitude may be "real" broadband's biggest challenge.