Nilsen, founder and chief technology officer of Java upstart NewMonics Inc., has invented a real-time technology that he believes is better at reining in increasingly complex applications than standard, plain-vanilla Java is.

The technology, in the form of a product called Perc, is essentially a vehicle for applying real-time smarts toward the tasks of navigating and managing complex network applications. Such apps will proliferate as the vaunted sphere of "convergence"-the combination of everything from television to voice to pagers and video gaming-begins to ride the same wire (or fiber-optic) loop.

"The biggest challenge is the complexity of the software," Nilsen said. "This has to do with the fact that there's a lot of code [required] to maintain connections in the first place and a lot of complexity in deciding how to reroute information when a signal is lost."

Nilsen learned that rule from his work with telecom providers. Such companies constitute perhaps the best proving ground for innovative technologies because they take their time about employing something new. As a result, once a telecom or datacom provider has bought in to a new engineering methodology, it's likely to be one that's more robust and battle-hardened than it would be in other applications arenas.

Nilsen made his homegrown Perc technology the basis of NewMonics, which he founded in 1996. Today, he serves as the company's chief technology officer, continuing to direct the architecture of it in forthcoming products. For the time being, however, NewMonics' flagship product will remain the real-time incarnation of Java, dubbed Perc.

"Perc solves a lot of the problems of dealing with software complexity," Nilsen said. "Perc is a special version of Java designed specifically for these large real-time systems where tight timing constraints have to be satisfied. So, if you need to run Java and still get latencies of less than 1 millisecond (ms), I think Perc provides a unique solution in the telecom space."

Nevertheless, according to Nilsen, the next generation of Internet architects will not find it easy to keep the engines of occasionally connected computing humming beyond the millennium. The biggest impediment will be the fact that smart phones, pagers and Web browsers connect and disconnect from the network at will.

Some companies have made stabs at solutions, such as Jini from Sun Microsystems and Universal Plug and Play from Microsoft. However, Nilsen believes those technologies may require too much processing power, memory and bandwidth to serve as a panacea over the long haul.

"The kinds of problems we see with telecommunications and telephone switching have to do with the fact that even the telephone system is not continually connected," Nilsen explained. "All the telecom equipment providers that we talk to say that it's very important to deal with the problem of wires and fiber-optic lines getting cut."

Indeed, in a real-time networked world, restoring service can sometimes be a nightmare, especially compared to the static telecom network of yore. "You have to find some alternative way to restore the communications channel-protecting the outage and working around it-within a very brief period of time," Nilsen said. In most cases, the maximum number bandied about is 50 ms.

"So it becomes a real-time programming problem-how do you detect the outage and how do you define the remedy for that situation within that very short time constraint of 50 ms," Nilsen said.

Surprisingly, Nilsen said, voice traffic will likely be a little harder to handle than data in tomorrow's networks. "Wireless data is a little more forgiving than voice, in terms of the real-time constraints, because you can tolerate a certain amount of burstiness in the packets," he explained.

With voice, on the other hand, the constraint is not so much timeliness, as it is correctness. One example that comes to mind involves cell phones that temporarily disconnect if a car or train enters a mountain tunnel. When they emerge, they may be "talking" to a different antenna. The difficulty then becomes ensuring that all the information that was supposed to have been received by the cell phone has actually been delivered to the device. For example, in an anomalous scenario, packets may be sent from one antenna, while acknowledgments that those packets have been received come from another antenna.

"There are some difficult issues there," Nilsen said. "The handoff has to be synchronized."

Looking at tomorrow's Net, Nilsen doesn't see a simple path to convergence. "We're not going to see a single solution that's going to solve everyone's needs in terms of how we're going to bring bandwidth into the home," he said. Whatever solutions evolve, Nilsen intends to remain in the game, using his undergraduate degree in computer science from the U. of Arizona and his PhD in physics from Brigham Young University as the basis for his technological forays.