The deal is similar to those cut by Tower Technologies in recent weeks with Compaq and Data General, said Tower marketing vice president Madison Cloutier. In the deals, the companies promote TowerJ server technology while TowerJ optimizes its product for the companies' servers.
The agreements highlight the increasing competition for products to run Java programs on servers.
However, IBM announced its own souped-up Java server solution two weeks earlier, and Compaq did likewise one week before Sun. Compaq announced benchmarks of its own Java technology running on Tru64 Unix as well as the deal to market TowerJ for Windows NT on Alpha chips. TowerJ's system is, functionally, a fourth take on this subject.
The deal Monday will involve optimizing TowerJ for HP-UX (HP's version of Unix), Windows NT, and Linux. In addition, Tower will announce version 3.0 of TowerJ on May 10, Cloutier said. The new version is better able to adjust to running Java programs that change midstream.
Compaq and IBM both are licensees of Sun's Java technology, but enhanced the software with their own improvements. All the companies focus on the Java virtual machine, the computer-within-a-computer that lets Java programs run unchanged on many different systems.
TowerJ, though, has taken a different approach. For one thing, it's independently developed software. The company is not legally allowed to say it's Java compliant or certified, Cloutier said, although technically the software complies with Java.
Another difference between TowerJ and the others is that it focuses on a more traditional method of changing the universal Java code into computer-specific instructions a chip can understand. Sun leans toward interpretation, translating the code line by line as the program executes, but Tower's technique is in the direction of compilation, which translates the code en masse in advance.
The different philosophies reflect different projections of how Java will be used on servers in the future, Cloutier said. Interpretation is better when a computer has to deal with Java code that changes often, but compilation works best when most of the code stays the same, he said.
"We're under the opinion that 80 percent of it is fixed, and only 20 percent is dynamic," he said. If the server-side Java environment should go toward a mix where 80 percent of the code is changing, though "HotSpot may beat us. But that's not what I see happening in the marketplace."
However, he said, TowerJ requires more manual tuning to get Java programs to work best, whereas HotSpot works to optimize programs automatically. The HotSpot approach, though, is a relatively untested, he said.