Skimping on Java May Prove Painful

I must confess to a certain intellectual perversity: a tendency, when listening to people who clearly know what they're talking about, to doubt their conclusions. In a previous column, I contradicted Paul A. Strassmann, who has written several books dense with econometrics that show that IT provides little or no boost in productivity. I argued that it does by improving the quality, if not the speed, of corporate decision-making. Of course, that was before Enron.

I find myself now doubting the conclusions of recent reports from Illuminata and Gartner, both of which say that end users can save money by developing most Java applications as lightweight Java 2 Standard Edition servlets rather than the more substantive applications that are possible using the full Java 2 Enterprise Edition. Servlets—smallish applications with limited business logic and little back-end connectivity—do the job quite well for most corporate applications, asserts James Governor, author of the Illuminata report "Java on the Cheap."

J2EE applications take longer to build; the tool kit itself is a lot more expensive, and most applications not only don't need the extra security and other features of J2EE, but they also run just fine on beefed-up Web servers like Apache Tomcat, which is stable, available and cheap, Governor writes. All true.

But recommending that people use servlets rather than J2EE for all but the most mission-critical applications ignores one major factor in the application development equation: fear.

Vital applications within an enterprise are no longer simply the mission-critical software without which the core business couldn't function. Now they include applications that end users believe they couldn't do their jobs without. Is e-mail mission-critical? It sure is now. Browsers? Yep. Word? You bet.

Think that even inconsequential Web apps aren't vital? Try shutting off instant messaging and you'll lose hundreds of man-hours of productivity just dealing with the complaints.

What's Mission-Critical?

Mission-critical is what users think it is. If you build a popular, useful Java app that sits on a $5,000 Wintel server but is "lightweight" because it doesn't launch stored procedures or remote procedure calls or connect through middleware to transaction processors, you'll get just as much grief when someone kicks out the power cord as you would if a half-million-dollar Unix box went out.

And you'll deserve it. Putting useful applications on risky platforms directly contradicts all of today's priorities in application development—most of which involve putting useful, user-friendly front ends on software whose business logic and connections sit on safe, manageable Unix boxes in a data center, where they can be maintained without having to be rebooted every couple of days.

You don't have to overbuild every application just to keep users from whining. Some applications really are inconsequential enough to run as servlets.

But if you're building an application that seriously helps your end users do their jobs, you're selling them and yourself short by not overbuilding it enough to make it reliable, scalable and easily manageable.

Contributing Factors

Governor doesn't agree with me. Neither does Gartner. Neither would anyone who did a rigid ROI analysis of servlets vs. J2EE. But ROI doesn't take into account the cost of users waiting for an unstable servlet, the cost of rebuilding a departmental application that's suddenly in demand throughout the organization, or the long-term cost of managing dozens of local servlets from an operations center that may or may not have the resources to deal with all of them.

There are ways to save lots of money and time doing Java development, and experts on that can tell you how. But I still can't escape the conviction that no matter how right they are, they're still missing the one nontechnical feature that can keep IT managers out of trouble: the ability to keep end users happy and productive, even if a certain amount of technical overkill is the price you have to pay to accomplish that.

Kevin Fogarty is editor at IT consultancy Illuminata Inc. in Nashua, N.H. Contact him at



EAI vs. Web Services

EAI vs. Web Services

Base: 100 CIOs (75 in the U.S., 25 in Europe)

Source: Merrill Lynch & Co., New York, March 2002

Special Report

New Tools, New Choices

Stories in this report:

Copyright © 2002 IDG Communications, Inc.

It’s time to break the ChatGPT habit
Shop Tech Products at Amazon