What about $500 PCs, you ask? Why buy a brain-dead thin-client device when PC prices are in free fall? Here's another chance for thin-client proponents to swing for the fences. First, while $500 PCs exist, most large organizations spend significantly more than $1,500 per new PC, or about twice the cost of a well-equipped thin-client device. Their money flows to high-end systems in the hope these computers will have a longer useful life. This strategy makes a lot of sense, because upgrading a PC is a time-consuming, costly exercise that almost always includes follow-up support calls.

More important, savvy organizations know that less than 20 percent of the true life-cycle cost of a PC is reflected in its initial cost. There's a mountain of evidence to support this assertion, as well as the corollary that thin clients save money. For example, a survey of 25 sites using thin-client technologies conducted earlier this year by Datapro concluded that on average, deploying thin-client devices cut support costs by more than 80 percent. If a thin client's purchase price were twice as high as a PC's, its cost of ownership would still be considerably less expensive.

But there is a catch or two. Even if you concede the economic advantages of thin clients, it's fair to question whether the approach represents technological progress. Many argue that it's a return to the old model of centralized host processing. You'd better be prepared to concede that argument and make the case that the previous way of doing business has benefits today. Thin-client deployments are most successful in organizations where centralized IT management can be embraced as more effective, not just more efficient. Even then, the systems have to deliver the applications users need with performance and functionality rivaling that of today's PCs.

These goals are not so easy to accomplish. While the momentum certainly is with thin clients, there are significant obstacles to successful implementation.

Perhaps the greatest challenge lies in thin-client computing's lack of a clear identity. A PC is a PC, no matter who manufactures it, and 95 percent of PCs use the Windows OS. It's clearly a commodity market around which a vast array of software applications has been built. Thin-client devices do not draw on such a common identity. There are Windows terminals, Java computers, network computers and even a new breed of ultraportable PDA devices. Instead of being characterized by a specific device architecture and OS, thin clients represent more of a new approach to computing. Don't think of the move so much in terms of replacing PCs with thin clients. Think of it as a new way to deploy applications across a range of devices, including PCs and, where appropriate, limited function thin-client computers.

So, is thin-client computing the kind of technology your shop should dive into headfirst, or does a wait-and-see approach make more sense? Even though a total shift to thin-client devices promises significant financial benefits, relatively few organizations are likely to replace their PCs entirely, both for technical and political reasons.

Somewhat more likely is the replacement of dumb terminals with thin clients, or the adoption of thin-client devices by workers who've never had desktop computers. In most organizations, thin-client computing will be linked to application rollouts or upgrades, and must therefore integrate with, rather than replace, the existing PC infrastructure. If you're planning a new intranet application, for example, you may find you can deploy it more quickly and cost-effectively by using Citrix Systems' MetaFrame and your current PCs. If you are deploying an interorganizational supply-chain management system, a thin Web/Java client may be the only practical option.

In short: If applications must be rolled out quickly, think thin. If network bandwidth is limited, think thin. If employees use a limited number of applications, think thin. If applications must be accessible by users whose desktop environments you cannot manage, think thin.