A great temptation these days is to hope that by implementing Web services, client PCs and devices will need to do less work. But it might mean the opposite.

---

by Andrew Binstock, principal analyst, Pacific Data Works LLC. Intel Corp.

What history tells us
A great temptation these days is to hope that by implementing Web services, client PCs and devices will need to do less work. Perhaps so much less that their normal upgrade cycle can be extended and IT budgets reduced thereby. This hope—conspicuously absent from the messages of Web services proponents—is based on the design of Web services that enables clients to have computation performed on remote Web servers.

Computing history and the nature of Web services argue against this position. Technologies such as Web services that allow clients to do more invariably necessitate greater computing power, not less. This seemingly obvious contention does not suggest itself immediately due to the quixotic drive for thinner, less powerful clients that touched the computing scene for several years in the mid-to-late 1990s.

The thin client movement
In those days, Microsoft Windows 3.1 was taking over the corporate desktop by storm, and IT managers felt considerable apprehension about the Microsoft and Intel dominance on the client side. Their concerns were based on what came to be known as total cost of ownership (TCO). The theory behind TCO was that PCs were no bargain if managers examined maintenance and support costs, rather than just the acquisition price. Viewed this way, the constant upgrade cycle of PCs and the need for regular technical support on software meant that corporations were spending far more on PCs than they previously thought. TCO was first widely touted by server vendors, notably Sun and IBM, but it was immediately picked up on by all major analyst organizations. Computations of the actual IT cost of a PC were remarkably wide-ranging depending on who performed the calculation and what costs were included.

The upshot of the debate about TCO was that both Microsoft and Intel initiated far-reaching programs to lower TCO. This included agreeing on and publishing the specifications for upcoming generations of PC hardware (published as PC98 in its first year, PC99 the following year, and so on) and a rededication by Microsoft to make its software more reliable and more user friendly.

Some vendors, however, proposed an entirely different solution: a scaled-down PC-like device that would download applications (and data) from a central server into local RAM and execute them there. This idea was a throw back to the days of dumb terminals, which were the dominant computing model through the 1980s. Dumb terminals were simply CRTs with keyboards. All processing, including the handling of individual keystrokes, was done by a mainframe to which the terminals were attached. In this set up, TCO was very low and predictable; the applications and the data resided on the mainframe and adding users simply meant getting them a terminal and a password.