Tuesday morning, I got up early for a D.C. power breakfast of the Washington Post and black coffee. The black coffee on an empty stomach helps counteract the nausea you get reading the latest goings-on on Capitol Hill, a subject the Post covers in excruciating detail.

I then strolled down 16th Street, past the White House and the Old Executive Office Building to the nondescript, generic federal-issue offices of the National Science Foundation. There, I went up to the 4th floor to find Steve Wolff, the Director of the Division of Networking and Communications Research and Infrastructure.

I was shown into a dark office. Steve strode over out of the shadows and introduced himself.

“I hate fluorescent lights. If you want, I can turn them on.”

I demurred and groped around for my notepad. There was just enough light coming in from the hall to take notes.

Steve Wolff took over the position of program manager from Dennis Jennings, the Irishman. Wolff helped formalize the NSFNET, supervising the shift from an interim 56 kbps backbone to a professionally managed T1, and later T3, network.

The solicitation for the T1 network had been won by a group that included the Michigan state network (Merit), IBM, and MCI. Later, a non-profit company was founded by MCI and IBM. This company, Advanced Network and Services (ANS), built a T3 nationwide network. ANS was inserted into the organizational chain with NSF and Merit.

When Steve Wolff and I talked in November, these organizational details had been ironed out and a nationwide T3 network, run in parallel with the older T1 network, was slowly becoming operational. Wolff was turning his attention to the next phase.

After numerous public hearings, Wolff was about to recommend to the National Science Board what the next phase of the network should look like. Needless to say, Wolff was not going to give me his formal recommendations before he told his bosses.Since I wasn’t trying to scoop Network World, we discussed more general issues.

Wolff explained the theory of network infrastructure as a pyramid. You could grow it by adding to the top or by broadening the base. Adding to the top was being done in areas like the national gigabit testbeds, intended to make the core run faster. Gigabit networks were still being tested, but it was evident that in some form they would soon allow researchers to start using higher bandwidth.

Adding to the base of the pyramid was equally important. Every year, NSF helped another 100 schools get on the Internet. This connections program was really quite simple. The school received U.S. $20,000, enough money to buy a router, pay for some telephone calls, and enslave part of a graduate student. After the initial U.S. $20,000, the school was on its own. To Wolff’s knowledge, not a single school had dropped off the network after the initial money ran out.

Seeding activities and leveraging investment are the fundamental strategies the NSF uses. They view their job as one of getting things started, not providing ongoing support. Most of the regional networks, for example, started off with exclusive or extensive NSF funding. By 1991, many of the regionals were self-sufficient. Even the core, NSFNET, received only a portion of its funds from the federal government.

My next stop that day was the Office of Technology Assessment where I raided the document room and even found a copy of a study I had contributed to several years ago, but had never seen the final results. I then headed out to Dulles Airport for my last stop of the trip, the Internet Engineering Task Force meeting in Santa Fe, New Mexico.