Moon’s Measure at Four-Tenths

Tonight at the observatory the meridian circle gave me the Moon’s position as 41° 17′ 28.6″, and my table of reduced elements demanded 41° 17′ 29″. The difference is four tenths of a second of arc. Nothing. Half the breadth of a hair at arm’s length. Yet in that flick of the pen, a number that existed - measured, real, earned from an hour of watching wire cross wire - was quietly replaced by a number that did not.

Rounding is structural, not cosmetic. Consider the circle of computation itself: the observation is a node, the table lookup an edge, the final reduction another edge. Between any two nodes I am free to carry either the exact value or the truncated one, and the topology permits both. But an Euler circuit requires every interior node to admit passage in and passage out with the degree unchanged. Truncation is an edge that eats one unit of degree and returns none. Do it once and the walk still closes. Do it at every node, as the tables compel - log to seven figures, sine to seven, the reduction again to seven - and the accumulated deficit of degree collects silently at the end of the path, where the orbit refuses to close and no map in the room will show why.

The bridge problem taught the lesson in geography; tonight it returned without it. Remove the river, remove the Moon even, keep only the arrows: measured value, rounded value, printed value. The orphan is always the true value. It has an outgoing edge - into the table, into the print - and no incoming edge, because nothing downstream ever reads it again. A quantity with no consumer is not preserved by the rounding; it is erased by it, and the erasure is symmetrical, half the time upward, half downward, so that across a thousand reductions the errors should cancel like even and odd degrees. Except when they do not, as in the 0.35 percent premium reported for expiration weeks - a small asymmetry, barely above the noise, whose entire substance may be nothing more than this accumulated truncation wearing the costume of structure. The market node may be a bridge that exists only on the stripped map, not on the ground.

I checked the tables by hand three times before the fire burned low. The four tenths are real; the 29″ is invented. Tomorrow I will recompute the lunar reduction carrying full double the printed precision at every intermediate node, and if the orbit still fails to close by a fixed residue, I will know the fault sits in the edges, not in the sky.