Nice little number, this one

Opinion: Smaller, neater, cheaper - the network computer has few drawbacks, argues Andrew Ward

Just over 100 years ago, the newly invented telephone was dismissed by people who thought messenger boys provided a sufficiently adequate service. Now, detractors are pouring scorn on the idea of the network computer, or NC - a simple computer that gets its power from the network to which it is connected, rather than carrying everything it needs internally.

But throughout the computer industry, massive investments are being made in network-centric computing. Are Sun, IBM and Oracle really all wrong? And if not, why not?

Some say the NC will not be much cheaper than the PC. That was recently disproved by Oracle, which demonstrated a network computer with a bill of materials of around $295 (pounds 200) - a fraction of the cost of the PC.

Scepticism is also voiced about the claims by Sun and Oracle for "near- zero administration costs". In fact, people are already implementing network- centric applications using PCs, so we know the administration costs for this computing model are low; replace the PC with the NC and they will be lower still.

There is also concern that the "newness" of the NC will result in teething problems. It will not. Technically, there is very little about the NC that is new. Yes, electronic devices do fail and the NC will be no exception, but its lack of moving parts and cooler running will give it a virtually indefinite lifetime - certainly far, far in excess of that of the average hot and noisy PC. And without a shadow of doubt, the NC will be smaller and more attractive than the ugly brutes littering our desktops today.

Network speed is another concern. But network-centric computing actually makes fewer demands on network bandwidth than existing PC-based applications. Database programs such as Access shunt vast amounts of data around networks, and massive applications such as Word 7 are often loaded from servers, creating huge bandwidth demand. NC programs will be much, much smaller.

And it is a nonsense to suggest that reliance on the network makes the NC somehow more exposed to failure. The reality today is that virtually all serious corporate use of PCs requires a network connection, so huge sums have already been invested in ensuring network reliability. Besides, there is nothing wrong with relying on a network; a telephone would not be much use without one, after all.

Another worry is a perceived lack of application software for the NC. This is an understandable concern, given the often considerable delays that beset development of new PC programs. But according to Ray Lane, of Oracle, "Sun's Java language will enable software suppliers to design and deploy mini-programs very quickly", so we can expect to see a greater variety of software coming to market more quickly.

The truth is that network-centric computing is not some idea for the future: it is here and now. "Web-enabled" applications, whose only requirement on the part of the client computer is to run a Web browser, are already up and running. Many tailor-made applications are in use now at organisations throughout the world, and off-the-shelf program suites such as Oracle's InterOffice are just around the corner.

Out in the real world, IT users are already convinced by the network- centric argument, which is why they are building these systems today. For the moment, the client software runs on PCs, but it is ready to take advantage of the much cheaper NC as soon as it becomes available.

Nevertheless, the NC will not take over completely from the PC, just as the telephone did not completely remove the need for messenger boys (we call them dispatch riders today). Nor will the PC disappear overnight. But as companies come to replace them, I believe they will increasingly turn to network computers. The PC will remain, but as an expensive and specialist tool, used by a minority.