Between 20 and 60 planes per hour -- that's all a runway can accept under instrument conditions, depending quite a lot on the skill of the approach controllers in the radar room.  Following a project meeting at NAFEC one day, three members of our research team stayed around to discuss these fundamental limitations with Henderson, our ATC client and mentor.

"If the arrival rate of air traffic goes higher than the runway capacity," I said at one point, "obviously something has to give.  You only have so many landing slots.  Stacks are a necessary evil."

Henderson nodded, which surprised me, considering our first conversation.

"Necessary, perhaps," interjected Claude Meyer, one of the computer programmers on our team.  "Why evil?"

"They cause too much delay," said Henderson.

Meyer feigned exasparation.  "You just conceded: stacks are the effect of capacity limits not a cause.  Delay is inevitable."

Henderson took chalk in hand.  "Say you only want a minute for spacing.  You put a guy in a stack, he's gotta stay there four minutes."  He wrote 1 - 4 = -3 on the blackboard.  You lost three minutes, maybe as many as three whole goddam landing slots.  That just backs up the incoming traffic even more.  Better off with dog-legs."

"Dog-legs have problems, too," said John McLaughlin, the mathematician, who worked with me on the Dual ILS project.  "First, they take up an awful lot of airspace," he said.  Henderson shrugged and handed McLaughlin the chalk.

"You have to give the guy a heading for the first segment," John said, sketching on the blackboard.  "You have to track the blip to the turn-point; you may be descending the guy through altitudes in the maneuvering area; and -- undoubtedly the most critical task -- you have to get the second turn transmitted and acknowledged at a precise time, otherwise you have a loose cannon on the deck."  McLaughlin finished speaking.

Before Henderson could respond, I spoke up with an observation.  "You've been telling us that controller workload can be horrendous.  With the stack, all you do is tell the guy to go to such-and-such a fix and hold.  If you never talk to him again, you still know exactly where he is -- and you don't have to eyeball his blip constantly either -- "

"Controller workload!" Henderson snapped.  He took a deep drag on his cigarette.  "You guys with your computers are going to solve all that, I read somewhere.  Are you giving up already, or what!"

"The way the problem is defined," I said, "bringing computers into the radar room will make controller workload worse!"  Oops.  My words had just tumbled out.  The room became quiet.  I could smell contract cancellation.

Henderson stubbed out his cigarette and smiled sarcastically.  "Ever since you guys came here," he said, "all I ever get is loony ideas for changing the ATC procedures.  First, it's 'speed control' -- slow 'em down, speed 'em up, whatever the hell."  Henderson held up his hand.  "No, wait.  It wasn't that -- first, it was some hare-brained scheme of holding airplanes on the ground at their departure ramp until you have a guaranteed landing slot open at the ETA, for crying out loud. Then, it was speed control, then -- "

Meyer interrupted.  "Speed control makes a big difference -- in both delivery error and in spacing on the common path.  You will see that with the simulations."

"If you'll let us try it," added McLaughlin.

Henderson continued, counting on his fingers.  "Then it was what you called 'multiple-choice' or 'sequence control' or some other damn thing.  Oh, yeah, and don't forget isosceles-damned-triangles in the sky!"  Henderson shook his head.  "Never anything about what the computer is going to do to help us poor slobs control airplanes better.  Oh, no.  And now -- and now, you're saying the computer is gonna make the problem worse?"

"You want me to explain?" I asked.

"I got all day."

"How many people does an approach controller have to talk to?" I asked rhetorically.  "Six, seven planes?  All right, there's the transition controller feeding him flight strips from the enroute center.  He's got one or more controllers right there in the radar room in adjacent sectors.  Up in the tower cab, by phone, is the local controller.  That's, say, 10 at various times.  Plenty, isn't it."

Henderson gestured for me to get on with my point.  I wrote the number 11 on the blackboard beside the word "computer."  Henderson stared at the board impassively.

Claude Meyer has a philosophical bent.  He had given much thought to this aspect of our assignment.  "Computers need data," he said.  "They don't make it for themselves, you know.  If anything, they reduce data.  But what you feed them better be good stuff.  Put in excrement and I don't have to tell you what comes out.  But, by God, the way things are right now, the only source for data is the controller."

"What data?" asked Henderson.

"Tracking data, mostly," I said.  "Where the planes actually are, moment by moment.  And altitude data."

Meyer grinned.  "Give us computer-types that good stuff and we can write algorithms 'til hell won't have it.  Woo-ee!  Pythagoras' Theorem, parametric in the time domain!"

Henderson was not amused.  He looked at his watch, a huge pilot's chronometer festooned with bezels and buttons.  The meeting was obviously over.  I followed him out of the conference room.

"It's true," I said.  "Look what happened in the enroute center.  A fiasco."

Henderson stopped in his tracks.  "Fiasco?"

"The FAA spent millions on those Librascope monsters," I said.  "They were supposed to print out the flight progress strips, right?  Well, the story I hear is that when weather is a factor and traffic gets tight, the first thing they do in the New York Center is call in their best controllers -- get 'em outta bed.  The second thing they do is, they stop talking to the damned computer.  And naturally, it stops talking to them."

Henderson appeared surprised.  I raised my voice.  "They go manual -- for hours, sometimes.  When calm is restored and there's plenty of time, only then do they start pumping stuff back into the machine.  But only then."

"Tracking data," repeated Meyer to Henderson's back.  "Let us automate the 'track-while-scan' function.  Then we'll get rid of your stacks."

"Not a chance."

There were many discussions like this over the next few months.  Each time I watched in vain for signs of acquiescence on the part of Henderson or any of our other ATC experts.

We were firmly reminded that our contract did not call for proposing changes to the system, only the simulation of ATC the way it already is.  Even in its infancy, computer technology was known to be poorly suited to that mentality.  Whether it's payroll or process control, you don't mindlessly "computerize" a manual system.

Meanwhile, something flickered inside my head.  A private flame kindled.  In a period of weeks, the conflagration consumed my mind.  The biggest idea in my life.