A funny thing is happening on the way to network computing. The model it’s supposed to replace–“client-server” computing–is teeming with innovation.
At the same time, client-server computing is becoming part of the environment that sucks in everything in its vicinity: the Internet. And we’re only beginning to gauge the impact.
If you’ve installed Real Audio on your personal computer, you’re part of this shift. Ditto if you use instant messaging software, such as America Online Instant Messenger or Yahoo Messenger, as well as a grab bag of other applications that handle a variety of chores.
More are springing up all the time. GuruNet pops up with information about any word on the computer screen; RUSure and Sherlock do comparison shopping; and News Alert provides updated headlines from MSNBC.
These applications would be utterly useless without the Internet. They aren’t Web browsers, but all use underlying standards on which browsers rely–TCP/IP (Transmission Control Protocol/Internet Protocol), the fundamental rules of the road for Net communications of all kinds, and HTTP (Hypertext Transport Protocol), the rules for communication between computers on the World Wide Web.
To understand this emerging trend, and why it’ll get even more interesting in the next few years, let’s recall some history.
Until the personal computer came along, computing largely consisted of huge mainframe machines (and smaller mini-computers) that dished out information to input-output devices, including dumb terminals–video screens, keyboards and little else. All of the computation took place on the mainframe or mini, which shared its processing among many terminals.
The combination of PCs and local-area networks brought about a new kind of computing: client-server. The PC was the client, and the servers were beefier computers–minis at first and then more powerful desktop computers–enabling the PCs to share files and printers.
More Top Picks Duster
Networks improved, and client PCs started running programs, or parts of programs, from the servers instead of the individual PCs. And as methods grew more sophisticated, it became clear that a smart way to handle computing was to let machines share processing tasks, letting each machine do chores for which it was best suited.
Before the advent of the Web, the linkage of individual PCs to client-servers was the essence of what became known as the “fat client”–generally Windows computers running bulky applications from companies like Microsoft, Novell and SAP (which got its start in the mini era). But with the Web and the Web browser, the nature of the client and the server changed.
Critics noted the enormous cost and complexity of running fat-client networks. It was hard enough to keep Windows running properly everywhere, but updating applications on thousands of machines was more than a chore. The solution, some said, was the Network Computer–the “thin client.”
Combine Web browsers, the Java programming language, powerful servers and ever-faster networks, the mantra went, and you could do everything you needed without having to run Windows at all. The server would do most of the processing, as in the old days of mainframe computers powering dumb terminals, but Web-based clients would be smart enough to handle enough of the job to make the combination workable.
The big advantage: By moving most of the complexity into the network, you could update software on the server and give the clients all the improvements simultaneously–and have an easier-to-administer network in the process.
The network computer hasn’t panned out very well, at least so far, though the notion still makes a lot of sense. Indeed, many of the applications we use all the time on the Web–search engines, shopping and more–are the essence of network computing.
But software developers aren’t ignoring the processing power in personal computers. And that’s one reason why Real Audio, ICQ, AIM and so many other applications are so popular and useful.
They differ from older client software in key ways. These programs are leaner than the older client-server applications, ranging from 200 kilobytes to a few megabytes. You don’t buy them in shrink-wrapped boxes. You download them from Web sites, and you typically don’t pay for them.
Because Windows is by far the dominant desktop operating system, and because desktop computers are the dominant device through which we connect to the Internet, the new breed of client applications are mostly Windows programs. Some developers are writing Macintosh versions, but few are doing or even considering Linux versions. So the emerging client-server marketplace tends to reinforce the Windows monopoly, at least for now.