|
Editor's Letter
June 2000
Universal it’s not
Intel, be merciful and let this gravely ill dog called USB die.
Maury Wright, Executive Editor
From its inception, USB was shortsighted in terms of its data rate and its PC-centric topology. Still, I can live with USB in a limited role; at least it lets my son shift between his various PC game peripherals without my help.
But in presentations at the recent WinHEC conference, Intel continued to pitch USB 2.0—the latest revision—even though the far more capable IEEE 1394 interface can also serve a desktop role (in addition to being the interface of choice in the digital living room and a potential home-networking technology).
USB 2.0 purports to push data rates up to 480 Mbits/sec (from the current 1.5 or 12). I won't argue that faster isn't better, but trust me, this issue isn't as simple as Intel and its puppets would have you believe.
First let's talk universality. USB was supposed to replace the rat's nest of cables behind our PCs. All USB devices use a common connector, PCs have several ports, and cheap hubs let you add ports in a tiered-star topology. Just plug a peripheral into any port, and you're set.
In reality, this universal scenario remains a pipe dream. Some products, such as network-related peripherals (for which USB wasn't intended) demand a USB channel all to themselves. Other bus-powered devices require careful placement in the USB hierarchy. Moreover, compatibility and driver problems persist.
But just wait for the horror of 2.0. To support both the faster data rates and USB 1.1 devices, the hub design and topology get so complicated that the typical user will have to hire a consultant to configure a system. If and when USB 2.0 ships, you'll have to connect 2.0 hubs directly to the PC and place 1.1 hubs downstream. If you put a 1.1 hub in front, the 2.0 devices won't even be able to sniff the faster rates. I guess Intel failed to tell us that they'd replace the mess behind our PCs with a bigger tangle on our desks. But hey, the connector is still universal.
|
Trust me, this issue isn’t as simple as Intel and its puppets would have you believe. |
|
By contrast, existing 1394 products work at up to 400 Mbits/sec in a peer-to-peer fashion, with or without a host PC. The new 1394b standard will introduce a new connector—after all, a jump from 400 to 1600 or 3200 Mbits/sec warrants a connector with better electrical specs. But get this: With only three different cable flavors, you'll be able to connect any new or old 1394 devices, and they'll all communicate at the highest speed the slower device supports. And new 1394b nodes will operate in high-speed clouds anywhere along the daisy-chained, hubless, hostless bus. Now, which standard is universal?
Topology isn't the only nightmare with USB 2.0. At WinHEC, Microsoft halfheartedly promised tentative plans to support USB 2.0 in Windows 2000 late this year and perhaps in Windows 9x next year. By contrast, 1394b requires no software update, and 1394 arguably boasts better software support already.
Since WinHEC prompted this column, I'll follow in the Microsoft tradition and put forth my own "call to action." Intel, be merciful and let this gravely ill dog die. And to anyone who agrees, let Intel know how you feel. Either contact Intel directly, or email us, and we'll make sure Intel hears what you have to say.
As Microsoft Technical Evangelist Joe Decuir put it at WinHEC, "1394 you can buy today, and it rocks."
|
