A model for Internet strategies.

Lately, we've been getting lots of visits from software executives eager to try out brand-new Internet strategies. (Now we know what it's like to live in New Hampshire during primary season.) Most of our visitors share a vision of the future--the very near future--that involves fundamental changes in the traditional model of personal, stand-alone computing. Whether the focus is on component-based network applications, shared documents, or the notorious $500 Internet appliance, there seems to be a consensus that a new computing architecture is emerging--and ignoring that trend is likely to be dangerous. Some further thoughts:

Is there a simple way to describe this new architecture?

In fact, that's a problem. It's not always clear how Web pages, Java applets, diskless PCs, corporate intranets, and all the other pieces of Internet-centric technology fit together into something that can be legitimately called an architecture.

Any suggestions?

Well, yes. If we're talking about distributed, interactive applications that run partly on desktop machines and partly on a remote host, there's already a pretty good term in use: client-server. The building blocks of Internet applications--HTML, TCP/IP, Java--may be different, but the basic model for Internet computing is arguably no different than traditional client-server computing. The application components happen to be connected by an Internet wire instead of a LAN wire, but that's not a big deal.

So why aren't client-server companies trying to play a bigger role in the Internet market?

A few are. It's no accident that the most vocal advocates for "thin client" devices are companies like Oracle, Sun, and IBM, who'd love to see a world where servers become the high-growth platform. But the awkward problem for just about every client-server company is that they have the wrong products. Most client-server development tools work only in relatively closed environments; thus, traditional client-server technologies have to be rebuilt from the ground up to work in the wide open spaces of the Internet. Again, it's no accident that Sun's Java is a brand-new language, or that IBM and Oracle had to acquire key Internet software components from outside the company.

Any chance that traditional client-server companies and Internet developers will converge?

That's the likely outcome. The client-server companies may not have the right technology, but they do have some incredibly valuable business models. They know how to price and support the tools they sell, they have a deep infrastructure of resellers and consultants, and they know how to find the big customers. And they've all learned at least one big secret that never seems to show up in the Internet strategy briefings we've seen lately.

And that secret is...?

The secret is that building large-scale distributed applications is much, much harder than it looks. Almost every traditional client-server company has learned the hard way that large projects often have severe performance problems, that it's tough to find competent developers, and that mixed computing environments can be as fragile as a house of cards. If the client-server world is any model for the Internet's future, then we're likely to see at least a decade of frustration and market chaos before a new distributed computing architecture become a reality.