Keeping a Measurement: Names, Dates and Files

The register runs on a single rule: a figure exists because it was written down, in a file that says what it is, on a date it was read. The rule is not about tidiness. It is about the difference between a measurement and a memory, which is the difference between a figure that can be checked and a figure that is just someone's word for it.

Why the file is half the measurement

Because a figure without its file is a figure without its context: no date, no method, no way to tell whether it was read this year or copied from a page that was itself copied. The file is where the context lives, and a named, dated file is the smallest possible archive that still deserves the name.

The housekeeping discipline is the same one that keeps any accumulation of files usable: a naming convention, a date on everything, and the occasional ordered cleanup rather than a heroic purge. The Mac side of that housekeeping is the subject of Headroom, which reads storage the way the desk reads figures: what is there, why it is there, and what can safely be let go. The longer-horizon version of the same discipline is written for preservation professionals at the Digital Preservation Coalition.

What a name has to carry

Enough to identify the file without opening it: the property measured, the material measured, and the date. A file called wrap-test is a file that will be opened for the wrong reason a hundred times; a file called film-gauge-cast-2mil-2026-09-09 is a file that can be found by someone who has never met the desk. The convention is cheap and the payoff is permanent.

The register's own version of the file is the standing table of measurements, one row per figure with the method and the date it was last read, kept in the spec table. A file name is the same discipline applied to a folder: the information needed to find the thing, on the outside of the thing.

What the date protects against

The silent revision, mostly. A figure updated in place without a date is a figure that has rewritten history, and the next reader has no way to know it changed. The dated file keeps both versions honest: the old figure and the new one are different files, and the difference between them is the record of the revision.

That is the same reason the register dates its rows rather than updating them in place, and why the batch record at the bench, described in what a batch code remembers, is written on the day the roll is opened rather than reconstructed later.

Where the file lives

Somewhere it can be found again, which is a duller and harder requirement than it sounds. A file that lives on the desktop of a laptop that dies is a measurement that died with it; a file that lives in a named folder, dated, is a measurement that outlasts the machine. The location is part of the record for the same reason the name is: a figure that cannot be retrieved is a figure that was not kept.

The desk's own rule is two copies and a folder that says what it holds, which is less a system than a habit, and which is exactly why it still works after the ambitious systems were abandoned.

What this page leaves out

Backup policy, which is a bigger subject than a file name, and which the desk handles the way any small operation does: two copies, one of them somewhere else. The point of this page is narrower: the figure on the page is only as durable as the file that holds it.

The register is itself a set of such files, which is why the discipline is stated rather than assumed. Every figure in the standing table is a line in a kept file, every article carries the date it was written, and the conventions described here are the ones the desk runs on because the alternative, a pile of figures remembered differently each season, is not a register at all. The file is the measurement's other half, and the measurement without it is a rumor with decimals.

A register built on kept files can be checked line by line, which is the only claim about its figures the desk will ever make. The rest is the reader's arithmetic.