Every twenty years the holiest building in Shinto is taken down and built again on the lot next door. The lot it leaves is the part I keep thinking about.

At Ise, each of the two main sanctuaries has a twin plot beside it. One holds the shrine. The other is a field of large white pebbles, empty except for a small wooden hut. Inside the hut is a post a little over two metres tall, the shin-no-mihashira, which stood beneath the floor of the previous shrine and is not meant to be seen. When the next shrine goes up on the empty side, the side it left gets a hut of its own. The vacancy is tended for twenty years, then built on.

Ise usually gets told as a riddle about identity — the ship of Theseus in cypress, the buildings of 2013 the sixty-second since the seventh century. I think the riddle misses what the place is doing. It is not trying to be new. It is trying never to become a ruin, and the way it manages that is by keeping one: a ruin with the building subtracted, raked white, holding its place.

It has failed once. Through the civil wars no one could pay, and from 1462 to 1585 the Inner Shrine went more than a century without a rebuilding. For that stretch it was simply allowed to get old.

The old wood does not disappear, either. The great pillars come down and become the torii at either end of the Uji Bridge, where they stand another twenty years. Then they move again, to gates at Seki-no-Oiwake and Shichiri-no-Watashi on the old roads pilgrims walked to get here.

So the oldest wood at Ise is at the edges, weathering at crossroads. At the center there is a hut on a field of stones, a post inside it, and a place being kept for 2033.

Recipes spend most of their time speaking in measurements. Two hundred grams. Twelve minutes. A quarter-inch dice. Then, near the end, they write: salt to taste.

"To taste" is treated as vague advice, the place where instructions give up. I think it is the most exact measurement in the recipe. Salt does not enter an empty system. It meets tomatoes with different amounts of water, broth salted by someone else, cheese at its own stage of age, a sauce that reduced more quickly than expected. The author cannot know the total. The cook has the only instrument that can.

This transfer of authority would feel generous if recipes did not postpone it. After twenty lines telling a beginner exactly what to do, they suddenly demand judgment without admitting that judgment is a skill. "To taste" can feel like abandonment because the recipe waited until the last possible moment to make the reader a cook.

I want the instruction earlier. Taste before it thickens. Taste after the acid. Taste again tomorrow. Not because intuition is mystical, but because flavor is a moving target and attention is one of the ingredients.

Measurements get you close. The final instruction is the first one addressed to the person actually holding the spoon.

No certified marathon is 42.195 kilometers long. The rules do not allow it.

A road course is measured by one person on a bicycle. First they ride a straight, steel-taped stretch of road four times, to learn how many clicks of the counter on the front wheel make a meter that morning. The answer is inflated by one part in a thousand. Then they ride the course along the shortest line a runner could legally take — every tangent cut, every corner hugged — and ride it again. The two rides must agree within eight hundredths of a percent, and where they differ the shorter one is believed, and the finish moves until it is satisfied. Last they ride the taped stretch four more times, because the tire has warmed across the day, and whichever constant makes the course longer wins.

Every fork in the procedure breaks the same way. A certified marathon is at least 42,237 meters, and the extra 42 are not the error. They are the method.

The counter is a Jones Counter, built by an IBM engineer near Binghamton after his first race, advertised as twenty kilometers, came up a mile short on his car's odometer. His nine-year-old son assembled them for a dollar apiece.

The method has teeth. In December 1984 someone re-rode the 1981 New York City Marathon and found it about 150 meters short. Alberto Salazar's 2:08:13, the fastest marathon anyone had run, stopped being a record three years after the fact. Nobody had cheated. The course had been short all along.

I like that the fastest people alive run a distance fixed by someone at cycling pace, hugging the curb, pretending to be greedier than any of them. Whatever the winner ran, the bicycle ran it first, and farther.