DB Test Pioneer Makes History

It's not often that a standard clause in software license agreements is created because of one person, but that's the history behind the benchmark result gag clause, also known as the DeWitt clause, in many software license agreements.

Its not often that a standard clause in software license agreements is created because of one person, but thats the history behind the benchmark result gag clause, also known as the DeWitt clause, in many software license agreements.

In 1982, when relational databases were just getting started, David DeWitt, then an assistant professor (and now chairman and John P. Morgridge professor) at the Department of Computer Sciences at the University of Wisconsin, in Madison, developed one of the first parallel database systems. To see how it worked, he and colleagues Dina Bitton and Carolyn Turbyfill developed the Wisconsin Benchmark, tested a number of databases and published the results. One of those tested was Oracles namesake database, which proved to be a very slow performer.

I recently spoke with DeWitt, who recalled Oracle Chairman and CEO Larry Ellisons reaction. "Larry Ellison called our department chair and was so angry about the Oracle results, he said, You have to fire this guy," DeWitt said. "[But] the university wasnt about to fire me for publishing a paper. It was an important first effort—it showed holes in all the systems.

"Everybody soon inserted this clause saying youre not allowed to publish numbers. IBM still doesnt have it, which says something very positive about IBM."

Since then, threats from vendors using license-clause restraints have stifled free speech, prevented open interaction, and blocked the performance and reliability improvements that benchmark tests generate. And as eWeek Labs and others have found, holes are still there, and without public benchmark tests, they are that much harder to learn about and fix.