Every Beam Since

The thing that happened over the following eleven weeks was not dramatic, which surprised Sabine and did not surprise Marchetti at all, and Rey had known it would go that way and had failed to explain it convincingly to either of them in advance.

Nobody evacuated anything. Nobody could have: there is nowhere in the belt to evacuate eleven habitats to, and the beams were over-section rather than under, and the honest engineering position from the first week was that the built environment was, if anything, slightly stronger than the drawings claimed.

What happened instead was that four hundred and eleven engineering firms across the belt received a notice, and every one of them stopped, and went and looked at their load models, and found in most cases that the discrepancy made no practical difference at all, and in nineteen cases that it did.

The nineteen were mostly drilling schedules: holes cut through webs at positions calculated from a nominal section, which in a beam nine millimetres over sat closer to the flange than intended and in six instances sat in a place that would have been rejected outright had anybody known the true geometry. None had failed. Two were within a factor of about one and a half of the load at which they might.

Palun's own habitat ring turned out to be fine, comfortably, on every span, which Rey learned in a four-line message and read on the gallery of the melt hall and then had to go and sit down about.

There was an inquiry, of course. It ran for two years and produced eleven volumes and its central finding was the one Marchetti had written in the galley on the nineteenth day: that no individual had been negligent, that the design mitigation had been adequate and had failed for reasons outside engineering, and that the industry should probably think rather harder about what happens to a twenty-year maintenance assumption when the company holding it stops existing.

Recommendation eleven, which is the one everybody quotes, says that no automated facility shall rely for its integrity on a report generated to a consumable, and it is known in the belt as the Paper Rule, and there are four generations of engineers now who have never heard of Hesper Foundry and could recite it word for word.

Sabine spent the eleven weeks in the melt hall producing the thing that made her name, and she did it badly at first and then extremely well, and the difference was that she stopped trying to do it alone.

The problem was reading the walls. She had a scan at a hundred and twenty microns and a theory that each band encoded a campaign's thermal map, and no way at all to get from one to the other, because the mapping from thermal field to tool path ran through eleven layers of a controller written ninety years ago by people who were extremely good and had no reason to explain themselves.

Marchetti found the answer in the bound release on deck two, in an appendix, on page nine hundred and six: the tool path generator was documented, completely, with worked examples, because the original engineers had assumed a maintenance technician might one day need to hand-generate a path with the controller down.

"Ninety years," he said, putting it on the table. "Ninety years ago a man sat down and wrote four pages of worked examples for a situation that he thought might arise, and it did not arise, and he retired, and he died, and nothing happened, and this week it saved us about nine months." He was very careful about how he said it and it did not entirely come off. "I would like it noted that I have changed my mind about several things."

After that it was work rather than discovery, and Sabine did four thousand one hundred and ninety-six of them, and Rey came down most evenings and read numbers off a gauge to her because two people doing it is four times faster than one, and Marchetti built the lookup and made it small enough to fit on the cheapest slate in the belt, which was his own idea and which he never once mentioned afterwards.

The curve, when it was finished, was the most boring-looking object Rey had ever been proud of. It rose from zero to nine and a bit millimetres over four thousand two hundred campaigns in a line so nearly straight that you had to be told there was structure in it, and Sabine printed it at four different scales before she found one that made the structure visible without making it look like an argument.

There was structure in it. There were fifteen flat spots where a campaign had aborted. There was a small step at campaign one thousand nine hundred and four, where something in the roll had been adjusted by a mechanism doing exactly what it should. And there was a very slight change of slope at campaign two thousand six hundred and eleven, which Sabine chased for a fortnight and finally matched to a feedstock supplier changing, forty years ago, on a hauler manifest, which nobody had told the foundry about because there was nobody to tell.

In the last week Rey did the thing she had been putting off since the second day, which was to walk the crew quarters properly and inventory them, because the assessment required it and because she had found reasons not to for eleven weeks.

Eleven cabins. All of them made up. All of them worked over, the figure running across the bulkheads and the desk surfaces and the inside faces of the lockers, at the density of campaign one, because the surfacing head had been through them exactly once, in the first campaign after the crew left, and never again, because the dressing routine only ran in occupied process spaces and a cabin is not a process space.

So the cabins carried the very first band, alone, at ninety years old, unrepeated. Rey stood in cabin four with her hand flat on a bulkhead that recorded the state of a foundry on a day when everything about it was still exactly right, and found that she was not able to be entirely professional about it and gave herself four minutes.

In cabin nine somebody had left a paperback in the locker, swollen and yellow, a detective novel in a language Marchetti could read and the others could not, and he sat down on the bunk and read the first page out loud in translation, badly, and Sabine laughed so hard she had to hold the door frame.

They found Ochoa's stamp in eleven places. On the release binding, on the printer rack, on two pumps, on the melt hall crane, on a hand-written card taped inside a panel on deck two that said, in block capitals, DO NOT ADJUST THE GAP WITHOUT READING P411, which was, Rey thought, the closest thing to a ghost the station had.

She photographed the card and sent it to the inquiry and it is reproduced in volume one, and there is a paragraph about it, and the paragraph is careful to point out that Ochoa had understood the risk perfectly and had left a warning at exactly the right place, and that the warning had been read by no one for eighty-six years because there was no one, and that this is not a failure of Ochoa.

Rey took nothing off the station. Sabine took the paperback, with permission, formally, on a chit. Marchetti took a photograph of the melt hall from the gallery, badly framed, with his own reflection in the rail, and had it printed on paper, at some expense, and hung it in a hallway in Palun where visitors regularly assumed it was a picture of frost.

They left on the ninety-first day. The capture ring released without being asked, exactly as it had turned to meet them, and the lights in the docking spine went down behind the lock as they cycled through, section by section, on the standard occupancy logic that every station on the belt uses.

Sabine watched it on the aft camera the whole way out. Nobody told her to stop and nobody joined her. At about four hundred metres the last interior light went off, and Hesper Foundry became a cold rectangle four kilometres on its long axis again, holding its attitude to a tenth of a degree, keeping its melt hall in shadow and its radiators edge-on.

"It has enough propellant for about eleven more years of that," said Sabine. "I worked it out. Eleven years, and then it starts to drift, and then in about forty the melt hall comes into the sun and everything in it cooks." "There is a proposal," said Marchetti. "Two proposals. One to refuel it and one to bring the plating home." "Which will happen?" "Neither, probably. The third proposal is to leave it, and the third proposal is cheapest, and I am afraid that I am going to support it."

Rey had expected to be angry about that and found, when it came, that she was not. She had been thinking for a fortnight about what the station actually was, and had settled on an answer she did not intend to put in any report: that Hesper had spent ninety years being conscientious in a room with nobody in it, and had produced, without meaning to and without any of the words she kept wanting to use, a complete and honest account of everything it had got wrong.

"They will not leave it," said Sabine. "Not once the plates are in the record. Somebody will go back." "In forty years." "Fine. In forty years." She was still watching the aft camera. "I will be sixty-six. That is not even unreasonable."

She did go back, as it happens, at sixty-eight, with a team of nineteen and a mandate and four haulers, and the melt hall plating is in eleven pieces in the long gallery at Palun where anybody can walk past it, and children put their hands on campaign three thousand and eight because it is at the right height, and there is a small unglamorous plaque that gives the whole of it in about two hundred words.

But that was later. On the ninety-first day there were three tired people on a small ship pulling away from a large dark rectangle, and the rectangle held its face to the sun and corrected on its cycle, and went on making beams, at nine and a bit millimetres over, into racks that nobody was coming for, and cutting the record of it into a wall in the dark, precisely, patiently, without any expectation whatsoever, exactly as designed.