Stern said, "What model is this?"

"La Roque," a technician said.

"But this model isn't accurate."

"Oh yes," the technician said, "it's entirely accurate. At least according to the latest architectural drawings they've given us."

"What architectural drawings?" Stern said.

At that, the technicians fell silent, worried looks on their faces. Now Stern saw there were other scale models: of Castelgard, and of the Monastery of Sainte-Mere. He saw large drawings on the walls. It was like an architect's office, he thought.

At that moment, Gordon stuck his head in the door. "David? Let's go."

He walked down the corridor with Gordon. Looking over his shoulder, he saw the technicians had turned the model on end and were carrying it through the door.

"What's that all about?" Stern said.

"Site-development study," Gordon said. "We do them for every project site. The idea is to define the immediate environment around the historical monument, so that the site itself is preserved for tourists and scholars. They study view lines, things like that."

"But why is that any of your business?" Stern said.

"It's absolutely our business," Gordon said. "We're going to spend millions before a site is fully restored. And we don't want it junked up with a shopping mall and a bunch of high-rise hotels. So we try to do larger site planning, see if we can get the local government to set guidelines." He looked at Stern. "Frankly, I never thought it was particularly interesting."

"And what about the transit room? What's going on there?"

"I'll show you."

The rubber floor of the transit site had been cleared of debris and cleaned. In the places where acid had eaten through the rubber, the flooring was being replaced by workmen on their hands and knees. Two of the glass shields were in place, and one was being inspected closely by a man wearing thick goggles and carrying an odd hooded light. But Stern was looking upward as the next big glass panels were swung in on overhead cranes from the second transit site, still being built.

"It's lucky we had that other transit site under construction," Gordon said to him. "Otherwise, it'd take us a week to get these glass panels down here. But panels were already here. All we have to do is move them over. Very lucky."

Stern still stared upward. He hadn't realized how large the shielding panels were. Suspended above him, the curved glass panels were easily ten feet high and fifteen feet wide, and almost two feet deep. They were carried in padded slings toward special mounting brackets in the floor below. "But," Gordon said, "we have no spares. We just have one full set."

"So?"

Gordon walked over to one of the glass panels, already standing in place. "Basically, you can think of these things as big glass hip flasks," Gordon said. "They're curved containers that fill from a hole at the top. And once we fill them with water, they're very heavy. About five tons each. The curve actually improves the strength. But it's the strength I'm worried about."

"Why?" Stern said.

"Come closer." Gordon ran his fingers over the surface of the glass. "See these little pits? These little grayish spots? They're small, so you'd never notice them unless you looked carefully. But they're flaws that weren't there before. I think the explosion blew tiny drops of hydrofluoric acid into the other room."

"And now the glass has been etched."

"Yes. Slightly. But if these pits have weakened the glass, then the shields may crack when they are filled with water and the glass is put under pressure. Or worse, the entire glass shield may shatter."

"And if it does?"

"Then we won't have full shielding around the site," Gordon said, looking directly at Stern. "In which case, we can't safely bring your friends back. They'd risk too many transcription errors."

Stern frowned. "Do you have a way to test the panels? See if they'll hold up?"

"Not really, no. We could stress-test one, if we were willing to risk breaking it, but since we have no spare panels, I won't do that. Instead, I'm doing a microscopic polarization visual inspect." He pointed to the technician in the corner, wearing goggles, going over the glass. "That test can pick up preexisting stress lines - which always exist in glass - and give us a rough idea of whether they'll break. And he's got a digital camera that is feeding the data points directly into the computer."

"You going to do a computer simulation?" Stern said.

"It'll be very crude," Gordon said. "Probably not worth doing, it's so crude. But I'll do it anyway."

"So what's the decision?"

"When to fill the panels."

"I don't understand."

"If we fill them now, and they hold up, then everything is probably fine. But you can't be sure. Because one of the tanks may have a weakness that will break only after a period of pressure. So that's an argument to fill all the tanks at the last minute."

"How fast can you fill them?"

"Pretty fast. We have a fire hose down here. But to minimize stress, you probably want to fill them slowly. In which case, it would take almost two hours to fill all nine shields."

"But don't you get field bucks starting two hours before?"

"Yes - if the control room is working right. But the control room equipment has been shut down for ten hours. Acid fumes have gotten up there. It may have affected the electronics. We don't know if it is working properly or not."

"I understand now," Stern said. "And each of the tanks is different."

"Right. Each one is different."

It was, Stern thought, a classic real-world scientific problem. Weighing risks, weighing uncertainties. Most people never understood that the majority of scientific problems took this form. Acid rain, global warming, environmental cleanup, cancer risks - these complex questions were always a balancing act, a judgment call. How good was the research data? How trustworthy were the scientists who had done the work? How reliable was the computer simulation? How significant were the future projections? These questions arose again and again. Certainly the media never bothered with the complexities, since they made bad headlines. As a result, people thought science was cut and dried, in a way that it never was. Even the most established concepts - like the idea that germs cause disease - were not as thoroughly proven as people believed.

And in this particular instance, a case directly involving the safety of his friends, Stern was faced with layers of uncertainty. It was uncertain whether the tanks were safe. It was uncertain whether the control room would give adequate warning. It was uncertain whether they should fill the tanks slowly now, or quickly later. They were going to have to make a judgment call. And lives depended on that call.

Gordon was staring at him. Waiting.

"Are any of the tanks unpitted?" Stern said.

"Yes. Four."

"Then let's fill those tanks now," Stern said. "And wait for the polarization analysis and the computer sim before filling the others."

Gordon nodded slowly. "Exactly what I think," he said.

Stern said, "What's your best guess? Are the other tanks okay, or not?"

"My best guess," Gordon said, "is that they are. But we'll know more in a couple of hours."