WAN bandwidth also is growing. For example, Lucent's WDM (wavelength-division multiplexing) enables five times as much data to travel over existing fiber optic cabling as before. This lets service providers add bandwidth rather quickly. There also have been increases in the carriage capacity of copper cable, now for analog modems up to x2/V.90, but soon perhaps for digital solutions such as xDSL.

But vast increases in bandwidth are not universally available. Have you been running xDSL lately? I didn't think so. Try it from South America or the Pacific Rim--you'll be lucky to get a moderately noisy analog line. Many rapid advances (like WDM) only benefit carrier backbones, not the whole end-to-end path.

Some of these breakthrough bandwidth solutions have economic consequences, such as massive carrier upgrades, that keep them from being widely provisioned, even in tech-savvy nations.

Sometimes a buyer just can't afford "unlimited" bandwidth. ISDN is a faster digital technology, but pricing in most places has made it unpalatable for consumers and businesses alike. And unlike resources such as CPUs, disks and LANs, WAN links are paid for monthly, not as depreciable, one-time capital expenditures.

One Asset That's Not Getting Cheaper Still, the trend exists: Costs for many resources commonly paid for by IT are diminishing. CPU MIPs, DASD and LAN bandwidth are all following a predictable trajectory (Moore's curve). Even WAN bandwidth is slowly getting less expensive and more available.

The only expense that's not declining is people. The cost of installing and supporting computing resources like CPUs, disks and networks is the only component of general costs that is not following a downward trend. In fact, it's the reverse. Wages are increasing, particularly as quality IT personnel become hot commodities. We've found, for example, that 50 percent of overall WAN costs likely will be for support overhead in another three to five years.

Because of these divergent trends, it's natural and prudent to want to trade costs in hardware, software, network bandwidth and other nonhuman resources for people. Spend more in one area and spend less in the other.

Invest in middleware to cut the time it takes people to program successful, deployable applications. Invest in overcapacity of many resources to reduce the time it takes people to minutely optimize the last few percentage points of performance from a CPU, a disk or a network. In fact, this methodology or ethos must pervade IT thinking from the top down and from the bottom up. I think it's starting--lately we see much work focused on total cost of ownership and return-on-investment calculations, with personnel costs prominently featured.

One thing's for sure: We will never have unlimited people. Well, I take that back. We'll just never have the budget for them. At least that's what I believe. I also believe this will prove to be a more operationally and strategically valuable belief than unlimited bandwidth.

Bruce Robertson is a program director with the META Group's Global Networking Strategies service. Send your comments on this column to him at Bruce.Robertson@metagroup.com.