Fractured iron warm to the palm

The chain link snapped at 84 tons. I had expected 100. The link came off the proofing machine at Bristol this morning, the fractured iron still warm to the palm, and the fracture face shows a flaw no calculation would have found - a seam of slag folded into the metal at the forge, invisible until the load found it.

Here is what I do not yet understand. The answer came quickly. The link failed; the fracture face tells me why; the forge can be corrected. Half a day, all told, from breakage to diagnosis. The question - what load should a link be proofed against - took me three weeks, and I am not certain the answer is right. I specified 100 tons because the worst cable tension under gale plus traffic plus shock suggested 60, and I doubled it. But doubling is a habit, not a calculation. The true question is: what is the heaviest load this bridge will ever carry, and can I name the conditions under which it arrives? I cannot yet. I can compute the steady tension. I cannot compute the gale arriving at the wrong moment, the crowd surging to one side, the temperature dropping the cable forty degrees. Those loads exist. I have not specified them.

The answers are cheap because the questions have already been paid for. Every fracture face I read this year was a question someone else failed to ask, and the iron kept the account. What worries me is the opposite case: the question I fail to ask, which the structure will answer for me, at cost, in public. The Great Eastern was designed for a route and a passenger trade I chose by instinct rather than by inquiry, and the instinct may prove the expensive part.

So the next action, decided this evening: before any further link is forged, I will write down every load condition I can name - gale, crowd, ice, shock, temperature - assign each a figure, however rough, and let the roughness itself show me which questions I have not yet learned to ask. The specification comes first. The fracture face is only ever the second draft.