A more mercantile problem with the vision of a ubiquitous lifestyle network, as seen by Suns Gosling, is the conflict of interest between providing platforms and providing content for those platforms. "In Japan and Europe," Gosling said, "youve got this interesting situation where the phone systems are regulated government bureaucracies, but they have strictly defined boundaries for their business. For instance, theyre not allowed to compete by creating games, so what happens is that they open their network. Anyone can create a game, and the games drive bandwidth consumption, and theyre motivated to let people do interesting things."Both Gosling and Adkins agree that another wireless infrastructure—GPS (Global Positioning System)—is on the verge of dramatically greater penetration into personal and enterprise applications. "When your car is on the network, and you combine that with GPS information, I dont understand why GPS receivers arent exploding," said Gosling of a technology whose obvious applications include everything from delivering organs for transplant procedures to delivering hot pizzas. Added Adkins, "Well start, based on location and presence, to deliver more information and capability to users based on their environment." Modes of user interaction will therefore have to adapt, Adkins said, to situations other than desktop and keyboard or handwriting tablet. "Youre going to see the 2003 Honda Accord with embedded speech technologies as part of the telematics, based on our embedded ViaVoice," Adkins said, adding that IBM is also working with Norways Opera Software ASA to develop a multimodal, text/voice Web browser based on the XHTML+Voice specification. Adkins credits some of IBMs rapid progress in these areas to open standards and open-source initiatives. "The beauty of our approach, being based on standards, is that we can repackage—from servers, to in-hand devices, now to an in-car device. We support a number of major standards, moving Linux to more embedded opportunities. ... Our approach across all IBM is to embrace the open movement, and much of what were doing is based on that. It gives us economic advantage." Suns Gage observed that much of the IT infrastructure is surprisingly unprepared for a sudden influx of new location-oriented data and applications. Questions such as "What points on this highway are within 100 yards of a natural gas pipeline?" have obvious bearing on homeland security concerns, as well as industrial operations, but dont fit into relational database models. "There are things that relational databases do very badly," said Gage, in Menlo Park, Calif. "A sense of things as parts of a whole—thats the desire for object databases, but theres an enormous investment in the schema for relational databases. "Why does SAP [AG] exist?" Gage asked rhetorically. "Theyve built data models of the world that more or less capture what an enterprise does. Thats why you pay them a billion dollars—because when you sit around and try to create data models, you realize thats really hard." At the same time, though, Gage asserted that developing and exploiting new data models is arguably the most vital challenge for enterprise IT. "Most of our business, in very large enterprises, is based on building and aggregating very large data models. Its a transitional mode and will be forever," Gage said. "Most databases make no provision for location codes, and when you have that information, you can do new things. The reason that the United States did so well in Afghanistan, after 10 years of work that began with Desert Storm, is that we finally had a unified terrain model, from people on the ground to B-52s in the air. Once that model is clear, and you have location information and other data, you can ask questions like Where is it going? When will it get to this point? Where was it when it experienced a drop of more than 1 meter?" Suns Papadopoulos agreed with this forecast of more information on individual items at every point in the supply chain: "Soon we will be adding miniature radio-frequency identification tags to all kinds of products, so we can instantly discover not only the price of the item but where and when it was made, how it was delivered, and a host of other useful data." Even where mobility isnt an issue, IBMs Adkins sees important opportunities from enriching interactions among devices in different locations. "One application that were developing with a large petroleum company involves gas pumps, refrigeration units and other devices that you find in a convenience store," Adkins said. "If the gas pumps are getting low, the store contacts a server. Stores were losing money with their high turnover and workers who didnt manage the environment, and they found that detecting the refrigerator being left open, for example, would quickly save them substantial amounts of money." Such diversity of devices, interacting in so many ways, could create an overwhelming burden of network management, Papadopoulos said. "In the same way the telephone industry had to look beyond switchboards and operators and move to automatic switching," said the Sun CTO, "the high-tech industry will have to find new ways to cope with increasing volume and complexity. "One way to do that," Papadopoulos said, "is with self-forming and self-healing networks—disparate devices finding each other across any kind of network, wired or wireless, without the need for centralized control, as in JXTA peer-to-peer technology. Clients and services forming and disbanding ad hoc networks without the need for human intervention, as with Jini network technology." All these interactions presuppose a body of code thats reliable enough to interact without human supervision. "I sure wish I felt more hopeful than I do" about software testing and other quality assurance disciplines, said Suns Gosling. "One of the things that Im pretty high on is systems that dont need as much testing because its in some sense correct by construction. If you peel the onion on the productivity figures for Java, a lot of that is because things just work." Microsofts Rashid said he also has hopes for reducing the burdens of testing through more reliable paths from specification to executable code. "There is, for example, in Java or the .Net Common Language Infrastructure, a kind of theorem prover that can establish that certain things will never happen," Rashid said. "Youve defined the byte code in such a way that a tiny number of rules can be analyzed in that way. What Im talking about is the next step—a much broader idea of the things that can be rules. For example, that if a certain lock is taken that it will always be released. The version of Windows thats going to come out two or three years from now will have drivers that have all gone through that discipline—its a huge advance." The hardest thing about transcribing a discussion like this is deciding who gets the last word—so hard that were not going to do it, but rather conclude with parallel points. The capacity of the network, rather than the power of the processor, is today the fastest-growing resource that drives the potential for dramatically new ways of using IT. "A single strand of fiber-optic cable can now carry more packets of data per second than even the fastest CPU can process. That has tremendous design implications," said Suns Papadopoulos. One of those implications is that much larger quantities of data now reside, in effect, close enough to one another that theres value in their combination and peril in their conflict: "Were going to be fixing these divergent, inaccurate data models of the world," said Suns Gage. "Theres a huge enterprise investment to be made." Technology Editor Peter Coffee can be reached at email@example.com.
Gosling, in Mountain View, Calif., expressed concern about the slower rates of emergence for new wireless applications in North America. "A lot of the companies sell the phones; they also sell the add-on services, and they want a monopoly on the services to your cell phone," he said.