People in the networking industry make predictions to sell their products, flog their stock, show everyone how smart they are, change public policies, and for any number of other reasons. The best predictions, the ones that add the most value, are those that come true unexpectedly. An example might be the folks at Citrix, who anticipated the demand for a server-based architecture with dependent, low-powered client machines, and who got a pile of money when Microsoft licensed it for future products.

The next best predictions are those that come true, but that can be seen as straightforward, if not trivial, extrapolations from the present: The number of Web sites will double every n months; the average size of a disk drive in a new computer will increase 40 percent; the cost of a MIPS will drop this much. Ho hum.

Then there are the predictions that are wrong, but their only real flaw is timing. Yes, it was the "Year of the LAN" for about five years before local networking became truly widespread in U.S. enterprises, but the idea was just too good to languish forever.

Finally, there are the really doggy predictions, the ones that didn't come true on time, and won't ever come to pass in a form resembling the original claim. I love to find these things-they're just so darned educational. Here are a few recent ones:

CABLE OPERATORS AS LOCAL COMPETITORS

With the passage of the 1996 Telecommunications deregulation legislation, numerous commentators predicted that a deregulated environment would result in cable companies competing with the incumbent local exchange carriers for voice services. Competing telephony services may some day arrive through the air or even over power lines, but cable companies won't be a big factor. A few cable operators, such as Cox Communications, may learn to leverage their fiber infrastructure by combining efforts with traditional telephony providers, or may even morph into fiber-based competitive access providers. But overall, the financial return from investing scarce cable operator capital in local telephony is always going to be sparse. Furthermore, as everyone knows, the cable service ethos is, "Slow, spotty, unreliable, and minimal." Can you imagine relying on the cable system to call 911 for an ambulance?

CABLE MODEMS AS BANDWIDTH PANACEA

The service shortcomings of the cable industry also militate against these providers becoming significant pathways for data services, such as Internet access. Surely, the key to survival for a retail ISP is to provide great service and support. Just how reliable and manageable will that Internet service be coming in over the cable infrastructure?

In the last year, cable companies and equipment vendors have announced orders for hundreds of thousands of cable modems, yet the number of actual installations at the end of 1997 was less than 100,000 (according to the cable providers themselves).

Though one cable system or another could service more than 90 percent of U. S. houses, only something like 10 percent of these houses could make two-way connections over the cable network. The rocky finances of the cable industry strike again: The cost to upgrade the cable network for two-way data transactions has been estimated to be anywhere from $500 to $2,000 per subscriber. To get a decent return on the investment, a high percentage of subscribers would have to sign up for the data service and stay signed up. None of your AOL-type "churn" is allowable here. Cable providers could use telephone lines and ordinary modems for the upstream link, but then they're selling the same story as satellite providers, such as DirectPC. Also, the installation and ongoing support burden would cut into their return on investment and play to their traditional service weakness.

The theoretical bandwidth of hybrid fiber-coax networks is seductive. Lots of big networking companies-Motorola, Bay Networks, 3Com/US Robotics, and Hewlett-Packard, among others-have worked hard to drive the cost of cable modems down. The unpleasant fact is that widely-available, sub-$50 residential Internet access with Ethernet-class throughput is five, if not 10, years away, and no more than a small fraction of such services will be provided by cable operators. That is, unless someone with money to burn-one of the incum-bent local exchange carriers or maybe Microsoft-takes them over.

NC AS PC REPLACEMENT

Unanswered questions about just what a Network Computer (NC) is have been around since the day their prime evangelist, Larry Ellison, first mentioned the concept. Is an NC a consumer set-top box that uses the TV as a display and lets you type e-mail by picking letters onscreen with a remote control? Is it a terminal replacement that gives you e-mail and Web browsing for the same cost as a dumb terminal? Is it a device that gives PC users all the functionality they're used to, but that costs less up front and lots less through its whole life?

If an NC is supposed to be the latter sort of device, it has failed definitively. E-mail, as long as it's text-oriented and not too fancy, is a simple application that won't tax a minimally powered system. Vanilla HTML displaying looks a lot like terminal emulation to a microprocessor. But if you start throwing in plug-ins, such as audio, video, and animation, or putting in some applets and gizmos that have to execute locally, we're no longer talking about cheap processors and less than 8Mbytes or 16Mbytes of RAM. In other words, you need all the components of a full bore, high-end PC, with the arguable exception of a $200 hard disk drive.

Larry's logorrhea got Intel and Microsoft all revved up about life cycle costs and, in recent months, practical ways to configure and manage PCs and desktop software have started to appear. NCs may replace X terminals and 3270 terminals, but replacement by NC is low on the list of potential worries for PC manufacturers.