Every so often I think about taking the bus. I’d like to be socially responsible, reducing our country’s dependence on foreign oil imports and lowering my personal contribution to atmospheric pollution.
Regrettably, my irregular schedule discourages the practice. After 5pm, the bus runs only once an hour, and if you’ve ever stood outside in a Minnesota bus stop in January, you understand the limits of social responsibility – and of centrally-controlled, shared resources.
Yet Larry Ellison, Scott McNealy, and Lou Gerstner all want me to embrace the network computer (NC) which, bus-like, will impose unnecessary and unwanted constraints on me.
The NC concept rests on false assumptions (FAs, to use the technical term). Among them:
FA #1: The NC and Thin Client are related: Nope. “Thin client” is a software architecture. Three-tier client/server architectures partition applications into independent processes handling presentation logic, “business logic” (a catch-all for a variety of different functions) and database services. You can implement thin-client applications on PCs. Heck, you can run everything on the same desktop PC and still have a thin client. It’s software, not hardware.
By the way, do you think “thin client” would be so appealing if they’d called the alternative a “thick client” or “muscular client”? And would it still be a “fat client” if it ran QNX or Linux instead of something flabby like Windows? I doubt it. This isn’t a dialog. It’s name calling.
FA #2: The NC and Java are the same thing: Well of course not. A PC can run Java code just fine. Not only that, but you can upgrade a PC with a better just-in-time Java compiler, language revisions, bug-fixes, and other stuff like that.
FA #3: The PC has huge hidden costs: This takes more than a one-paragraph tirade. A hint: everything written on the subject is a cost analysis, not a cost/benefit analysis, not a cost comparison with alternatives, and not the only analysis that matters: a cost/benefit comparison with alternatives that provide equivalent functionality.
FA #4: Java is portable: Wait a few years. You’ll have new models of the NC with enhanced functionality. You’ll have proprietary extensions. You’ll have competing, incompatible NC architectures. Suppliers differentiate themselves from each other, because if they don’t they’re selling commodities, and commodities have razor-thin margins.
FA #5: The NC has no hidden costs: Once you buy into NCs you’ll need more bandwidth, bigger servers, and more sophisticated network, server, and applications management.
Why do you think IS installs applications on PC hard disks instead of servers now? It’s the bandwidth needed to download DLLs from servers. Download a DLL, download a Java application, what’s the difference? Java doesn’t magically shrink the size and complexity of a word processor or spreadsheet.
And since the NC is simply a paperweight when the network, server, or application crashes, you need better management of all three. Clue: who wants to sell you the servers and management software?
Here’s why the NC won’t succeed. People bought PCs specifically to gain independence from control-oriented IS organizations. The PC freed employees to do what they chose, not what IS decided they should do.
The claimed benefit of the NC is its greatest hidden cost: With it, IS morphs back into the god of software, arrogantly dictating what end-users can and can’t use. If the NC succeeds, end-users will simply re-invent the disconnected PC or independent departmental LAN, so they can do what they want, instead of what IS lets them.
Technology doesn’t only follow business requirements. Sometimes it creates business opportunities, or leads to unexpected business results.
Adoption of personal computer technology in the 1980s led inevitably to employee empowerment. Look at the arguments in favor of the NC and you’ll discover a thinly veiled attempt to disenfranchise employees again.
As is true so often when you peel the onion a bit, it’s enough to make you cry.