ther processor," he asserts, explaining that it has do with shared memory buses and snoopy caches. "To get 100 processors, you have to go to another architecture," Mitchell says.
Right now, Spring is mostly the basis for research. The situation at Brown University is fairly typical. "We're looking at object-oriented approaches to deal with replication in distributed systems," says Tom Doeppner, research associate professor at Brown. The advantages eventually would be increased fault tolerance and performance, with dynamic load-balancing and server switching.
"We're taking advantage of subcontracts," says Doeppner. "The client thinks that it's invoking operations on the object directly, but our subcontract steps in and encodes the fact that the object may exist in several places." So when the client invokes a method, the subcontract finds a suitable server, caches the information, and continues to use the server until it's not a suitable choice anymore.
Purdue University
is working on garbage-collection systems. "We have a conservative garbage collector for C and C++," says Vince Russo, assistant professor at Purdue. "And we just started working on network security for object systems," which would work at the subcontract level.
The University of California at Santa Cruz is looking into performance instrumentation for Spring. "We were very impressed [with its performance] on the door calls," says associate professor Darrell Long. "We're developing a performance monitor."
Most of this work is in the early stages, but all the university projects show promise of commercial use. It's too bad, the professors universally agree, that Spring is not headed toward becoming a commercial OS.