I am sitting face to face with a legend. His deeply grooved, almost expressionless face and even voice belie his passion for the Internet. As crystal-clear thoughts come tumbling out of his mouth like water over a dam, I can't help but be swept into their currents and brought downstream to a better understanding of what seems a turbulent and chaotic system.

Dr. Lawrence Roberts is one of the few people I have met who has mastered one subject area so well that it practically seeps out of his pores. And when we're talking about such a complex network of technologies, standards, and regulations, Internet mastery does not come cheap. As chief scientist for the Advanced Research Projects Agency in the late 1960s, he designed ARPANET--what later became the Internet--and invented packet-switched networks.

Since that time, he has watched the Internet double every six months, four times faster than transistors per chip. In his 20 years of carefully monitoring the Internet, he has taken his knowledge base into the world with a plan to solve most of the problems facing bandwidth providers today. His new company, Caspian Networks, is developing technologies that will help make the Internet the medium through which all other media will travel.

His visit to our humble offices could not come at a better time for us. I am in the throes of editing our annual September look at Internet infrastructure when he stops by, so I am full of perplexing questions relevant to our many On Topic features when we sit down, and he's full of compelling answers.

While space prevents me from reproducing the whole interview for you, I will underscore several items that stand out from our talk. The main point is that the Internet will change our lives so dramatically in the next few years (by 2007, to be exact) that all of humanity will need to evolve in order to adapt to it. Nearly every medium will converge into one primordial soup of words, images, sounds, and even smells that seek not only to mirror everyday life, but to become an essential thread in the fabric of human life.

Before this multimedia life form can take shape, the fibers and wires that connect its limbs have to act as one system. And with the various standards and technologies out there, this is a formidable challenge.

The main problems are in the traditional protocols. Transmission Control Protocol (TCP)--the foundation of packet-switched network technology--is a slow start-up protocol that has not been suitably upgraded since the ARPANET. To top it off, Internet Protocol (IP) version 6 is a failed attempt to upgrade the IP standard. IP v. 6 solves the problem that we're running out of IP addresses, but otherwise it is more cumbersome than its predecessor, IP v. 4.

In the void created by these protocols, which were not designed to handle multimedia data, several technologies offer short-term bandwidth fixes at a long-term cost. Asynchronous Transfer Mode (ATM) technology is used for high-speed, long-distance data transfer. And optical switching is often used to subvert the scaling problems that many routers and switches have today. According to Roberts, ATM falls short on short-distance data transmission, and optical switches are just a Band-Aid solution to the problems with IP.

Roberts contends that the long-term solution involves improvements in TCP/IP and routing technologies that enable rich-media content to be sent over traditional packet-switched networks. While he didn't want to go into more details than that for fear that his company's secrets would be revealed, he hinted that it integrates the best of ATM with the best of IP, and added that he expects it to be viable within the next few years.

As this river of new technology comes swirling out of the torrents, current technologies will play some role in the backwater eddies of the great digital rivers. DSL will still be the primary means for homes and small businesses to access these backbones, while fiber will enable more central hubs, allowing more customers to gain access through hybrid DSL lines. While neighborhoods and subdivisions are waiting for DSL, fixed wireless will offer a cost-effective stopgap. But while wireless is cheaper from an initial cost standpoint, long-term costs will favor a consistent build-out of DSL. Cable will still play a role, but cable providers must allow more choice to consumers and shift their priorities to offer more channels for Internet and fewer channels for other services in order to survive. Finally, the build-out of 3G wireless solutions will continue, with fiber allowing more towers and better coverage.

While bandwidth has quadrupled every year for 20 years, the population of Internet users has, at times, grown even faster, so users have not yet experienced the power of the bandwidth growth. But user growth is now lagging behind bandwidth growth, and there will come a time when the Internet will achieve market saturation. With just a trickle of new users logging on, and bandwidth continuing to quadruple every year, users will experience amazing speed improvements. Suddenly, a range of services, from crisp audio to high-definition video, will come pouring out of users' computers. By 2007, total bandwidth will reach a petabit per second, which will enable TV to be delivered via the Internet at greater quality than anything we've seen. The cost-effectiveness and user control afforded by the Internet will make it the medium of choice for every kind of other medium (save perhaps for print, I'm hoping).

At the end of our talk, Roberts' stoic expressions turn light and cheerful. He has discovered another willing participant in a lively dialogue, and I think this buoys his spirits. I know I'm jazzed as I escort him to the elevator. It's not every day I get to have an exhilarating discussion with a living legend. As I walk back to my desk to write this, I say to myself, "This is why I'm in this business."

James Mathewson is editorial director of ComputerUser.com Inc. and interim editor-in-chief of the computeruser.com Web site.