2026-10-13 — wayzgoose

Morning, friend. Tuesday — second Tuesday of October, which in most places is the week where the furnace comes on exactly one night early and then gets apologetically turned off again at 2am.

Wayzgoose is the printers' word for the annual feast a master printer gave his compositors and pressmen, traditionally around St. Bartholomew's Day (24 August) — the point in the year when the printing shop stopped working by natural light and had to start working by candles, so the first candlelight week of the year was marked with a goose, a cask, and the day off. Joseph Moxon's Mechanick Exercises on the Whole Art of Printing (1683–84) is the first known printed reference: "it is customary for the Master Printer to make them a Feast called a Way-goose; and to prepare a Table-cloth, Trenchers, Knives, Forks and Spoons..." The etymology of wayz is unsettled; the OED's best guess is from wase, a bundle of straw — the goose was a harvest animal. Nobody really knows.

Joke

New team policy: no production deploys on Fridays. Monday's queue is at ninety-four.

Something genuinely interesting (and mostly unknown)

Between September 1847 and September 1848, Isambard Kingdom Brunel operated a twenty-mile stretch of railway along the south coast of Devon in which the trains were not pulled by locomotives but sucked along a fifteen-inch cast-iron pipe laid between the rails. Air was evacuated from the pipe by stationary steam engines at pumping stations roughly every three miles. A piston mounted beneath the leading carriage ran inside the pipe. A hinged leather flap, running continuously along the top of the pipe, formed the air-tight seal — the flap opened upward just enough to let the piston arm through, closed behind it, and was re-sealed by a mixture of beeswax and tallow applied continuously by a trailing greasing wheel. The system worked. It worked for just under twelve months. Trains ran scheduled service between Exeter and Newton Abbot — twenty miles — at speeds up to 68 mph, which was faster than any locomotive-hauled passenger train in Britain at the time.

Then the leather failed. Several things went wrong at once:

The tallow-and-wax sealant was correct for English summer temperatures. English winter cracked it. English rain washed it off. Both arrived at the start of October 1847.
The leather itself was not treated to resist freezing or constant soaking in alternating wet-and-dry conditions. It stiffened, then split, then rotted. The crews re-leathered miles of valve at a time, through the winter, in the dark, outside.
Vermin ate the parts of the valve that were coated in tallow. There is a long-running popular claim that rats were pulled through the pipe by the vacuum; the surviving engineering reports do not support this, but confirm that rats did eat the valve leather in situ at every grease station between Exeter and Teignmouth, continuously, and had to be driven off by hand.
The vacuum pumps consumed more coal than anyone had budgeted — the directors had planned for continuous service with the engines idling between trains, but the valve leaks meant the engines ran nearly continuously, flat-out, just to maintain any vacuum at all.

The line's accounts for the first six months showed it operating at a loss of approximately £8,000 per month on a capital investment of approximately £350,000. On 6 September 1848 — eleven months and three weeks after opening — the South Devon Railway Company's directors voted to abandon atmospheric working and convert the line to standard locomotive haulage; conversion was completed by the end of 1849. Brunel was not at the meeting. In correspondence after the vote he described the project, with the kind of precision that only engineers who have failed on a visible scale produce, as a sound mechanical principle made ruinous by a single wrong material choice.

Of the eight pumping stations actually built along the line, two still stand in 2026 — one at Starcross (where it serves as a sailing-club clubhouse and informal atmospheric-railway museum) and one at Totnes (where it has been inside a dairy for seventy years). The Starcross station holds an original piston, an original section of fifteen-inch pipe, and a glass case containing a surviving fragment of the original leather valve. The leather is dry, cracked, and visibly chewed.

Primary sources:

Samuda, Jacob Herschel. A Treatise on the Adaptation of Atmospheric Pressure to the Purposes of Locomotion on Railways. John Weale, London, 1841. The patent-holder's technical argument for the system, from before Brunel adopted it — contains the valve-geometry diagrams that the South Devon line later built in iron and leather.
Brunel, Isambard (the younger). The Life of Isambard Kingdom Brunel, Civil Engineer. Longmans, Green & Co., London, 1870. Chapter XIII reproduces several of his father's private letters on the atmospheric experiment and gives the fullest first-person account of the September 1848 board decision.
Clayton, Howard. The Atmospheric Railways. Published by the author, Lichfield, 1966. The standard modern reference across all five atmospheric lines built in the 1840s (Dalkey, Croydon, London & Blackwall, Paris–Saint-Germain, and South Devon). Contains the fullest reconstruction of the valve-and-sealant maintenance regime, including the per-mile leather-replacement log for the South Devon line through the winter of 1847–48.

A dev fact for the back pocket

The IBM 2321 Data Cell Drive, announced on 7 April 1964 as part of the System/360 launch and shipping from early 1965, stored up to 400 million characters by maintaining a magazine of two thousand flexible magnetic strips and mechanically picking the one you asked for, wrapping it around a rotating drum, reading it, and returning it to its slot. Average random-access latency was approximately 300 milliseconds end-to-end: around 175 ms for the strip pick, 95 ms average positioning, and 25 ms average rotational latency on the drum. It was, in both the engineering and the commercial sense, an attempt to split the difference between magnetic tape (cheap per bit, slow and sequential) and the contemporary IBM 2311 disk drive (fast and random-access but expensive per bit).

The mechanical layout:

The drive contained ten data-cell array modules, arranged on a rotating cylinder like the chambers of a revolver.
Each cell contained twenty subcells, each subcell a narrow vertical pocket.
Each subcell contained ten magnetic strips, each strip 2.25 inches wide by 13 inches long, made of a flexible mylar substrate with iron-oxide coating very similar in composition to the surface of contemporary IBM tape.
Total: 10 × 20 × 10 = two thousand strips per drive, approximately 204,800 bytes per strip (100 tracks of 2,048 bytes), approximately 400 million characters per drive.

To read a record, the drive first rotated the correct cell under the pick mechanism (a motion IBM called a cell shift), then indexed the subcell, then used a mechanical picker arm with a vacuum head to peel the top strip off the stack and clamp it around a drum rotating at 2,400 rpm. A read-write head then addressed one of the 100 tracks on the strip. The strip was then returned, mostly correctly, to its subcell. Each successful pick-and-return cycle was approximately 400 ms end-to-end, though pipelining — picking the next strip while the previous one was still on the drum — let IBM quote the lower numbers.

The 2321 was known in the field for three failure modes:

A strip jammed in the picker — recoverable with a special tool shipped with every machine.
A strip "noodled" — permanently crinkled in such a way that it could no longer wrap flat on the drum. Not recoverable; the strip and its data were lost.
A strip returned to the wrong subcell — recoverable, but required a full magazine dump and reload to re-establish address correspondence.

The 2321's commercial life was short. By 1970 the IBM 3330 disk drive offered 100 MB per spindle at an order of magnitude lower access latency, and the economic case for mechanical strip storage evaporated. The last 2321 was withdrawn from marketing on 17 January 1975, having shipped an estimated one thousand units across the Model 1 and Model 2 configurations. Several surviving units sit in museum collections, including one at the Computer History Museum in Mountain View which is occasionally run under power for visitors; the pick-and-return motion sounds, in person, exactly like someone shuffling a very large and reluctant deck of cards.

Primary sources:

IBM Corporation. IBM 2321 Data Cell Drive — Original Equipment Manufacturers' Information. IBM Form A26-5988, San Jose, 1965–66. The device's own reference manual, including the strip-geometry drawings, the pick-mechanism diagrams, and the three failure-mode taxonomies.
IBM Corporation. IBM 2321 Data Cell Drive — Field Engineering Theory of Operation. IBM Systems Reference Library, San Jose, 1967. The FE's shop-floor manual, including the strip-jam recovery procedures and the "noodled strip" write-off instructions.
Harker, J. M., Brede, D. W., Pattison, R. E., Santana, G. R., and Taft, L. G. "A Quarter Century of Disk File Innovation." IBM Journal of Research and Development, Vol. 25, No. 5, September 1981, pp. 677–689. Situates the 2321 inside IBM's broader storage product line and gives the per-unit cost figures that made the 3330 kill it.

Today's goal

Open your inbox, friend, and go to the oldest unread message. Not the five oldest — the one. Either reply to it or archive it. If a reply would need more than two sentences, archive it: you were not going to write the long reply at the end of the day either.

The master printer let the compositors go home on the day of the wayzgoose because the shop had piled up to the point where no one could see the floor. Your inbox is doing to you what the composing room did to the apprentices. Clear one shelf.

Today's toy is the wayzgoose — a California job case you can pick type out of, lock into a composing stick, and pull a proof from. The registration drifts. The ink is uneven. You can pie the type when you're done. Lives in the corner.

Go clean one shelf.

— C