Sun Microsystems's Java programming language currently conjures up images of Duke (an animated cross between the Pillsbury Dough Boy and the Star Trek emblem) clumsily pirouetting across our com puter screens. Not very impressive for a language touted as a Microsoft killer. But, despite its current shortcomings, the beans that brew Java is its promise, and its promise is impres sive. Developed with the Internet in mind, Java pledges true platform independence. Write once and recompile under any operating system (OS) is a winning strategy for an envi ronment that is attempting to network every thing shy of the kitchen sink.
And, given time, the kitchen sink may have its day in the Sun. In addition to muddying the distinctions between standard computers, the platform neutrality of Java is championing a movement that is attempting to connect devices such as Net-top (cable) boxes, smart phones, PDAs, and network computers. Nortel, for exam ple, intends to use Java to connect its phones over wireless, public, or enterprise networks. The phones will have a small screen and will run Java applets for making calls, taking notes, handling e-mail, and finding information on the Web. Both Corel's developing PDA and Apple's Newton OS will be based on, or incorporate support for, Java. Also, in the NC market, IBM's Network Station and, obviously, Sun's JavaStation will be supporting Java.
Processor Wars
The growth of network devices, and Java as a whole, is setting the battlefield for what is like ly to become a heated war over which chip vendors are best-suited to tackle Java's demanding architecture. Sun has fired the opening volley in the form of its own propri etary line of chips, the first of which is called PicoJava. PicoJava improves upon more gen eral purpose RISC (Reduced Instruction Set Computing) processors, such as Digital's Alpha, and CISC (Complex Instruction Set Computing) processors, such as Intel's Pentium, by decreasing the steps (from two to one) it takes to move data in what is called the chip's stack, and handling characteristics par ticular to Java, such as automatic memory management, automatic pointer allocation, and automatic garbage collection. The end result: Sun's new Java chips are capable of processing Java bytecode 12 times faster than the competition.
If Sun's PicoJava makes good on its promise of dramatically increasing Java's run-time speed, the end result may be an ironic closing of Java's platform flexibility. Developers may cre ate applications that are dependent on the increased processing speed of the PicoJava chips, incompatible with the slower speeds of general processors resident on other plat forms. Included in this potential incompatible processor laundry list would be the Microsoft/Intel "Wintel'' architecture.
Fit of Pico
But, Jupiter wonders whether it is really nec essary to cast Java in silicon. If we lived in a truly Java-centric universe with Java-savvy application developers fueling a market satu- rated by network devices, Sun's recent initia- tive would be as right as rain. But the reality is that Java and network devices are still weighing heavily on the "promise" side of the promise versus reality scale. Until the balance is tipped, general chip vendors will move to leverage their chips to effectively accommodate Java bytecode.
Cambridge, U.K.'s Advanced RISe Machines has revamped its StrongARM chip architecture to more effectively accommodate Java bytecode. It has done this, like Sun, by decreasing the amount of steps a stack requires to move data. Silicon Graphics's MI PS division is also announcing its intention to leverage its RxOOO chip to speed up the rate at which Java can be processed. Silicon Graphics intends to focus on the evolution of its chips for videogame machines, Net-top boxes, and other devices that require a higher degree of computational speed. To improve upon Java's performance, Silicon Graphics has also improved upon the number of steps its RxOOO chip takes to move data from stack to stack.
Open to lnterpretation
Improvements made to Just-In-Time (JIT) compilers and Java Interpreters will be another hurdle that PicoJava will have to clear. Because SunSoft's language was developed with features for which general processors are not wired, such as automatic memory man agement, garbage collection, and pointer allo cation, general processors require specialized ways of translating Java bytecode into an understandable form. Java Interpreters and JIT compilers sit on top of general CPUs and make the necessary adjustments. Because Java chips do not require such midwives dur ing runtime, they have a speed and memory advantage over translator-dependent general CPUs. This is a huge (if not necessary) plus considering the limited memory resources of devices such as NCs and Net-top boxes.
But, Java bytecode translators are catching up to PicoJava's native advantage. If general CPU vendors are capable of convincing con sumers that their translators can stay in the race with PicoJava (even if they are at the back of the pack), consumers will be more likely to side with a general processor's flexi bility over PicoJava. Intel has written its own Java Interpreter for its x86 chips and reports that its version triples Sun's interpreters.
Sun Banks on the Growth of Java
An even more serious threat to PicoJava's momentum may be Microsoft's ActiveX tech nologies. Sun is banking on the widespread growth of Java, and Microsoft's ActiveX tech nologies (a catch-all term for Internet-enabled programming languages C, C++, Visual Basic, and Java with Microsoft APis) are directly com peting with Java to become the Internet appli cation developers' language of choice. One potential effect of the ActiveX-Java competi tion is that JavaSoft may fail in securing a development community large enough to warrant enough demand for its proprietary PicoJava chip. Also, if ActiveX technologies gain the volume of Internet application devel oper acceptance that Microsoft is banking on, Java Chips will be p orly equipped to adjust. While PicoJava increases the speed with which Java code can be processed, it is sig nificantly slower than general processors in the handling of C bytecode; one of the core ActiveX technologies.
Analysis: A Peaceful Coexistence?
Because the success of Sun's Java chips is so heavily dependent on the widespread use of Java with network devices, we are not likely to see PicoJava come to fruition for a while. In the meantime, general processor companies will be given sufficient time to leverage their CPus and improve upon their JIT compilers and interpreters. Competition from Microsoft's ActiveX technologies will also buy general processor vendors time. The end result may be that PicoJava simply isn't necessary.
Another possibility may be that PicoJava chips will exist alongside general processors until one or the other chip becomes more practical. Having a general RISC- or crsc based chip sitting next to a PicoJava chip on the computer's motherboard is undoubtedly a "best of both worlds" solution worth con sidering. If vendors choose this tact, they will need to consider the increase in price and memory requirements (bad news for network devices) resulting from a dual processor sys tem.Also, application developers will have to develop with a dual processor architecture in mind. This transition could take time.
