Inhaling network computers (first appeared in InfoWorld)

Here’s a scary fact: Kofi Annan, the new Secretary General of the United Nations, went to the same college I did, although his stint at Macalester preceded mine by a dozen years. Maybe he needs help managing his technology. Do you think?

At Mac in the early seventies, the grass always looked greener under the other guy’s Gro-lux. It wasn’t, of course – it smoked just the same. (I, of course, didn’t inhale, already planning my future in politics, and Mr. Annan left Mac before weed became easy to come by there.)

Based on reader responses to my debate with Nick Petreley over the Network Computer (the RL/NP NC DB), it’s clear our industry isn’t immune from greener-grass disease, either. NCs sit across the fence. They don’t exist yet, so they possess magical properties that solve every problem ever ascribed to the PC.

Reality won’t be so thrilling, of course, because Java is slow and prototype NCs run like sludge. But don’t worry: Java chips, just-in-time compilers, or some other yet-to-be-invented innovation will take care of that. Or else Digital will take Nick’s advice and reposition its Alpha chip as an NC processor to bring Java up to PC performance levels.

But I’m tired of arguing, so instead I’m going to help all you NC advocates make the system succeed in your organization. Proper migration to any new architecture requires careful planning to minimize disruption. Here’s the program:

Step 1: Migrate to all Java applications: Every NC advocate has explained that NCs won’t replace all PCs – just some of them. Since electronic document sharing will continue to be important when you add NCs, you’ll need to use the same word processors and spreadsheets on both NCs and PCs. Right now that means converting to Corel Office across the board.

This is a low-risk move anyway, so jump right in. If you compare Corel’s price and licensing, you’ll see immediate economies of between 400 and 1000% even if you change your mind about installing NCs.

Go for it.

Step 2: Inventory all departmentally developed applications and write replacements in Java: Whether you know it or not, end-users have built a bunch of small systems using tools like Access and Paradox, or even plain old Excel with some macros. Power users build them, but lots of just-plain-ordinary users rely on them every day so you have to replace them with something that will run on NCs. This shouldn’t take more than a few staff-years of effort, and it will be worth it in the long run. Especially when you figure that since the NC doesn’t have anything equivalent to Access, departments will come to you again for these small applications instead of developing them on their own.

Step 3: Convert to SMTP/POP/IMAP e-mail: Well you’re not going to be able to run cc:Mail, MS Exchange or Groupwise on an NC, are you? You need to use an e-mail system you can access through a browser, and that means using the same e-mail technology used in the Internet – not a bad idea when you get right down to it.

Step 4: Beef up your LAN with a high-speed backbone, switching hubs throughout, and high-performance network management: You’re going to need the bandwidth, since you’re going to be downloading all of your applications through the LAN, and you’ll need the network management because any network outage will have a much bigger impact than before.

Step 5: Clear out most of your Netware servers and convert them to Web servers: You could use Netware’s Intranet products, but they’re not renowned for their great TCP/IP performance, and nobody has been promoting NCs that use Novell’s IPX protocol.

Now you’re ready to install NCs around the company. And it should be worth it, so long as real NCs have the same characteristics as the slideware we’ve been basing our plans on.