Rumors are flying that IBM and Sun Microsystems are codeveloping a next-generation version of Unix that will tightly incorporate Java. The goal: Deliver a scalable Java application environment and simplify Unix use.
Without referring to this project at all, Sun Solaris marketing director Brian Croll said, "Five years out, you will see Solaris be a Java server engine. The fusion of Java and Solaris is going to give the capability to run a huge application on the server side, and it will just scream."
Deloitte & Touche's High Technology Group recently released its "1998 Software Industry Annual Report," a review of major developments in the software industry in 1997 and predictions for 1998.
Based on 1997 performance (and a Zona Research study), the report sees Java wading into battle as a non-proprietary alternative to Windows, slowly absorbing more software developers with its design. Zona projects that the Java tool market will triple (from 8 million to 80 million) from 1997 to 2000.
The High Technology Group's managing director Mark A. Evans said it wasn't as easy to spot a technology winner this time around. "This past year, the industry's rules were radically changed by bold new products, hard-ball marketing, and high-stakes litigation," he said. "Our [Report] takes a close look at the industry, how it's changed, and the challenges that await it -- and provides insight into how it has affected virtually every aspect of our daily lives."
Other findings indicate that business spending will be twice the projected volume of consumer online spending by 2002; online gaming networks should breathe new life (and billion in revenue) into the entertainment software industry; and the Asian market crisis means postponed hardware purchases and infrastructure reductions.
For a copy of the "1998 Software Industry Annual Report," send an e-mail request to hightech@dttus.com.
At Intel's lead, software for the next-year 64-bit Merced chip (operating systems, compilers, and so on) will be Java-lite, centering on C++ instead. (Compilers and the OS will be important factors in leveraging the architecture of the Merced chip.)
Intel itself, and HP, Microsoft, Metaware, and Edinburgh Portable Compilers are all developing Merced-capable C++ compilers -- not an easy task as the Merced architecture uses both the speculation and the predication techniques (which seem to be mutually exclusive).
To take advantage of both, a compiler has to be able to let the many execution units skip unnecessary decision branches and avoid memory latency.