W e b w o r d

Vol. 1, No. 5: SEPTEMBER-OCTOBER 1997



Charles Petrie

Stanford Center for Design Research


Knowing that my computer science orientation is AI and agents, you might wonder what I would care, or know, about intranets. As it turns out, the subject is not only more important, but also more interesting than I first suspected.

AI is all about computational representation. If you and I represent things the same way, we can share more with less. That's why we don't need so much bandwidth to talk. Because we share (or think we do) so many of the same concepts and speak (or think we do) the same language, we can convey lots of information with just a few words. Try to recall the last time you listened to geek talk with a fresh ear.


Remember when no MIS manager dared buy anything but an IBM mainframe? Or when Unix became ubiquitous, killing many better but proprietary operating systems? Or how about the success of DOS? What happened?

Companies like to standardize. Purchasing likes to buy a lot of one thing. IS likes to maintain one platform and homogenize software. So how do they settle on that one thing? (Hint: It's not necessarily the best thing.) The decisions are due mostly to marketing and the bandwagon effect.

In marketing, the perceived common denominator wins, whether lowest or not. So companies put in a bunch of Wintel boxes running something common and well-supported. Linux loses. Win95 wins. What will happen to Unix and NT is the crux of the matter here.

By establishing an intranet, a company can standardize on TCP/IP technology within the network, allowing everyone to access the Internet while restricting incoming communications. With an intranet, everyone in the organization can use the same networking technology instead of a proprietary one.

A lot of networking software companies underestimated the force of the urge to standardize/homogenize. Novell, one of the top networking software providers, was left hanging on to proprietary software when many of its customers moved to intranets. Ultimately, standardizing displaces a lot of installed bases, a hundred-year flood over proprietary barriers.

IS likes intranets because it means IS staff has one homogenous technology to support, on which they can build a multicompany career. Purchasing likes them because they increase competition. Users like them because they can go through them out on the Internet. And management likes intranets because they can build extranets.


There are only three catches: security, e-commerce, and network services. These are the important technical issues and the fuel for a super-heated industry.

Security and e-commerce are fairly well understood applications, and the need for common standards has already led to frantic competition among software companies. Network services is not so well understood and the competition is just flaring up. We know we need directory services. Almost certainly we need file distribution. Do we need consistency management of updates for replicated files? It's a tough problem that was never really solved by the database people. So how are we going to solve it for intranets?

And then there's the real bear--transaction management. This is the ability to roll back a transaction if crucial parts of it fail. In a distributed transaction, this means that for every participating application the software engineer must identify the correct success/failure points. A lot of folks are pretending we only need this functionality in very specialized applications. I bet they're right. The problem is just too hard, so people are going to have to design failure-tolerant systems.

But all of these services, to the extent they are offered, need to be standardized. And everyone wants their standard to become the industry standard. Own the standard, own the market. That's the dream of every vendor. Microsoft did it for operating systems. Now they want to do it for the Internet and intranets.

Opposing the Microsoft empire is the virtual machine consortium - a loose alliance betting that IS folks will go for homogenization of services based on a Java platform-independent standard, rather than homogenization based on NT servers. And the consortium has a chance. The majority of network servers are not NT machines.

Own the Standards, Own the network

The dream is that Netscape, IBM, Novell, Sun, and others will present IS vendors with real network software choices, all guaranteed to run on a variety of platforms and all with a common technical base--the Java virtual machine--along with a variety of intervendor standards (for example, Lightweight Directory Access Protocol (LDAP), Java Electronic Commerce Framework (JECF)) for network services, e-commerce, and security. So now there is a race to be first, and not necessarily with the most.

The kicker is that there is at least one dark horse here. Deutsche Telekom's T-Online is the world's biggest intranet, if you think of subscribers and suppliers as a virtual company. While US companies are squabbling, they already have a set of network services in place, including a functioning electronic commerce system. And they are forcing US ISPs to work with them. Given that they are moving so aggressively, possibly in partnership with France Telecom, they could put de facto standards in place that US companies will have to live with.

So the question is, who will win the intranets standards battle? And what will the resulting standards be? Think about your own dissatisfaction with your company's proprietary network software and its crummy e-mail and file sharing. I know you're not happy.

Own the standards, own the network.

Aren't you glad you're here at the beginning to help choose?

Other URLs for this Column

Interview with Eric Schmidt

France Telecom is a big investor in Intershop - http://www.intershop.com.
Deutsche Telekom is Intershop's biggest customer. Deutsche Telekom, France Telcom, and Sprint have formed a "Global One" alliance.

For a discussion of T-Online, see "Internet in Europe," in IEEE Internet Computing, Vol. 1, No. 1, pp. 84-85, 1997. Also available at http://www.computer.org/internet/9701/europe9701.htm.

For an explanation of relevant acronyms and issues, see Miro Benda's column "Middleware: Any Client, Any Server," in IEEE Internet Computing, Vol. 1, No. 4, pp. 94-96, 1997. Also available at http://neumann.computer.org/ic/books/ic1997/pdf/w4094.pdf.

Back to IC Online

Copyright (c) 1997 Institute of Electrical and Electronics Engineers, Inc., All rights reserved.