2026-08-24 — fandangle

Morning, friend. Monday. The last one of August, and the calendar is starting to move with that particular Monday-in-late-August posture where the machinery is running but everyone has one eye out the window at the cottonwoods, which have started to turn from the tops. The week hasn't been told the season is over. The season, quietly, has already begun packing.

(Fandangle — English noun, chiefly dialect, an ornamental contrivance of no practical purpose, a piece of foolery got up with more work than the result deserves. From Spanish fandango — the eighteenth-century courtship dance from Andalusia, first attested in print in 1712 in the anonymous Carta al Reverendo Padre Fray Martín Sarmiento — with an English diminutive ending accreted onto it in the mid-nineteenth century by way of theatrical and music-hall slang, where a "fandangle" was a piece of stage business too elaborate for its role. The OED cites its earliest printed English use in Thomas Sternberg's The Dialect and Provincialisms of Northamptonshire (London: John Russell Smith, 1851), where it is glossed as "any fantastic ornament; a piece of gaudy show; a bauble." The word survived in American English via the mid-twentieth-century revival of dialect vocabulary in the funny pages: Al Capp's Li'l Abner used it repeatedly through the 1940s, always for a contraption Marryin' Sam or one of the Yokum cousins had built for no known reason. A fandangle is defined by the ratio of effort to purpose. The ratio is bad. The fandangle is the point.)

Joke

The comment matches the code once, at commit time, by accident.

Something genuinely interesting (and mostly unknown)

In the summer of 1954, in the limestone chambers of Luray Caverns in the Blue Ridge foothills of Page County, Virginia — a commercial show cave that had been open to the public since it was located by tinsmith Andrew J. Campbell on the afternoon of 13 August 1878 — a mathematician by the name of Leland W. Sprinkle began walking through the caverns with a rubber-tipped mallet and a chromatic tuner, striking every stalactite he could reach and listening carefully to what came back. Sprinkle, then in his fifties, worked as a mathematician for the Department of Defense in Arlington and had come to the caverns because his young son had bumped his head on a stalactite during a tour and produced, by accident, a clean musical tone. Over the next three years Sprinkle identified thirty-seven stalactites, distributed across some three and a half acres of the cavern's principal chambers, whose natural pitches fell close enough to the equal-tempered chromatic scale that a light and patient grinding with a small stone tool could bring them onto true. He wired each of the thirty-seven to a solenoid-driven rubber mallet, ran the wiring back along the tour path to a central console fitted with an ordinary organ keyboard, and on 3 June 1957 turned it on. The instrument he had built is called the Great Stalacpipe Organ. It has been in continuous playable service ever since, and by Guinness's measure — the area over which its resonating elements are distributed — it is the largest musical instrument in the world.

The mechanism is worth walking through. The keyboard console sits in a small alcove off the Cathedral chamber, thirty metres or so below the parking lot; the organist plays it as they would any three-manual chamber organ, and each key closes a low-voltage circuit that runs, via braided copper cable strung along the ceilings of the intervening rooms, to one of the thirty-seven electromechanical strikers scattered through the Titania's Veil, the Cathedral, the Dream Lake, and half a dozen smaller chambers. The strikers themselves are simple: a coil, a plunger, a rubber tip, mounted in a small steel bracket fixed by dental-cement adhesive to the limestone shoulder of the target stalactite so as to strike it laterally an inch or so below the base of the ceiling. When the key is pressed the plunger fires, the rubber contacts the mineral, and the whole three-and-a-half-acre chamber system rings — because a limestone stalactite is a tuning fork attached to a resonating vault. Some tones ring for a full seven seconds before dying out. Chords sustain, overlap, and cross with themselves. The room is the instrument as much as the stalactites are.

Sprinkle patented the mechanism. United States Patent 3,049,958, "Musical Instrument," filed 30 September 1957, granted to Leland W. Sprinkle of Springfield, Virginia on 21 August 1962 and assigned by him to Luray Caverns Corporation, describes in five pages of small type the striker circuit, the mounting bracket, the mallet material selection (natural gum rubber over a soft-iron core, selected against seven other rubber compounds for its acoustic response against dense calcium carbonate), and the tuning procedure — a light abrasive removal of no more than three millimetres of surface material per grinding pass, tested with a calibrated tuning fork between passes, applied only to those stalactites whose natural pitch fell within a semitone of a chromatic target. Sprinkle explicitly noted in the patent's claims section that the invention "may be applied to any suitable subterranean deposition of speleothems in a resonating chamber," which is one of the more optimistic sentences in the history of the United States Patent Office. No second Stalacpipe has ever been built.

The organ is played today by a small rotation of tour guides trained in-house at the caverns; the standard tour piece is A Mighty Fortress Is Our God (Luther, harmonised by J. S. Bach), which fits the pitch range of the thirty-seven struck notes and rings through the Cathedral chamber for a little over three minutes. Weddings are held at the console. A recording engineer from RCA Victor visited in 1958 and produced an LP called Sounds of the Great Stalacpipe Organ which sold, per the caverns' own count, thirty-four thousand copies by 1961 and is now a collector's item of the outsider-music record trade. Sprinkle died in 1990 at ninety, having spent the last thirty-three years of his life quietly tuning the instrument every few years as the limestone deposition changed the pitch of the target stalactites, a maintenance schedule he had described in the patent as "occasional, and not more onerous than the tuning of a fine harpsichord." His son, the one who had bumped his head, occasionally still visits.

The instrument is a pure fandangle. It is the largest musical instrument on earth, by area; it plays one Lutheran hymn well and one Bach cantata passably; and it exists because a mathematician noticed his child had made a note by accident and spent three years of evenings and weekends turning a cave into a chromatic scale. The ratio of effort to purpose is bad. The instrument is the point.

A dev fact for the back pocket

On the morning of Friday 10 April 1981, at Launch Complex 39A at the Kennedy Space Center, the countdown for the first flight of the Space Shuttle Columbia — STS-1, commanded by John W. Young with pilot Robert L. Crippen — was halted at T-minus 20 minutes because the four IBM AP-101 general-purpose computers running the primary flight software, and the fifth AP-101 running the independently written backup flight software, could not synchronise with each other. The primary software (PASS — Primary Avionics Software System) had been written by IBM Federal Systems Division in Houston in the language HAL/S, and had been running in the four primary GPCs for the previous thirty hours of pre-launch testing. The backup software (BFS — Backup Flight System), written to the same functional specification but independently by Rockwell International's Autonetics division in the same HAL/S, had been powered up two hours before the halt for its normal pre-launch sync-up. It could not read valid data from the primary system. Young and Crippen were on their backs on the pad, strapped in, breathing pure oxygen, listening to Launch Control try to figure out why. The launch was scrubbed. It flew successfully forty-eight hours later, on Sunday 12 April 1981 — the twentieth anniversary, deliberately timed, of Yuri Gagarin's flight — after the ground team had power-cycled the primary GPCs and brought the backup up fresh against a clean primary. Behind the sync failure was a race condition whose probability of occurring during any given initialisation was later calculated at approximately one in sixty-seven.

The specific mechanism was a timing skew between the primary and backup systems introduced during the primary's initial power-up sequence, thirty hours before the halt. The primary AP-101s ran on a 40-Hz major-cycle timing frame — every 25 milliseconds, the four primaries exchanged state, voted on their outputs by strict bit-for-bit majority, and started the next cycle. The backup ran on the same 40-Hz frame, but with its cycle boundary offset by 20 milliseconds from the primaries so that its independent computation could read the primary system's outputs mid-cycle and vote on their reasonableness end-to-end. The backup's initialisation routine assumed that when it came online, the primary would be on cycle boundary N, and the backup could therefore align to boundary N + 0.5. But because of a specific software artefact of the primary's own initialisation — a two-cycle "settle" period during which the four primaries synchronised themselves before beginning the numbered cycle count — the primary's first numbered cycle could, with a probability estimated by the post-mortem team at one in sixty-seven, be cycle 1 rather than cycle 0. If the backup came up during that specific case, it aligned to 0.5 while the primary was on 1.5, and every subsequent inter-system communication was one cycle out of phase. The two systems then agreed on nothing.

The bug had never been seen in fourteen months of integrated testing at Downey and at the Cape. It had never been seen in the parallel Shuttle Avionics Integration Laboratory (the SAIL, at Johnson Space Center in Houston), which had run the identical software in the identical hardware configuration for the equivalent of some four thousand shuttle missions of ground time. It had never been seen because the two-cycle settle period had been introduced in a routine software patch to the primary six weeks before launch, and the SAIL had been updated to that patch, but the odds of hitting the one-in-sixty-seven state combined with the odds of the backup coming up in that particular quarter-second-wide vulnerability window meant that in all the SAIL runs since the patch, the state had never obtained. On the actual pad, on the actual morning, it did.

The definitive post-mortem is John R. Garman's paper "The 'Bug' Heard 'Round the World", in ACM SIGSOFT Software Engineering Notes, vol. 6, no. 5, October 1981, pp. 3–10. Garman was Deputy Chief of the Spacecraft Software Division at NASA's Johnson Space Center and had personally been in Launch Control during the halt. His paper is nine pages long, extraordinarily clear, and closes with the sentence: "The very existence of the backup system saved the day, but the very existence of the backup system had also, in a sense, caused the problem." That is a sentence a person writing about redundant systems should read once a year.

The lesson, thirty-five years on, has not aged. Every distributed system with a fresh replica that must synchronise against a running primary contains, somewhere in its initialisation code path, a variant of the STS-1 bug. The specific window is small. The consequences of hitting it are not. The mitigation, then and now, is the same: bring the primary down and back up clean before you bring the replica up. On 12 April 1981 that took two days and cost the United States Government approximately five million dollars in delay and standby costs on Complex 39A. Your Redis replica costs less and takes less time. Do it anyway.

Primary sources:

• Garman, John R. "The 'Bug' Heard 'Round the World." ACM SIGSOFT Software Engineering Notes, vol. 6, no. 5, October 1981, pp. 3–10. The primary and definitive post-mortem, written by the NASA JSC official responsible for the flight-software configuration control that had accepted the six-weeks-out patch, with the full timing diagram of the race condition and the SAIL simulator statistics on why it had not been caught in ground test.
• Cooper, A. E., and W. T. Chow. "Development of On-Board Space Shuttle Software." IBM Journal of Research and Development, vol. 20, no. 4, July 1976, pp. 371–384. The IBM Federal Systems Division's account of the primary software architecture, six years before the flight, including the four-computer vote scheme and the intended backup interface. Reading this and Garman's paper together shows the design as intended and the way in which reality did not conform.
• Tomayko, James E. Computers in Spaceflight: The NASA Experience. NASA Contractor Report 182505, Washington, D.C.: NASA, March 1988. Chapter 4 (pp. 87–171) is the authoritative institutional history of the Shuttle GPC and its software, with a full section on the STS-1 sync bug and its subsequent handling in the STS-2 and STS-3 software loads.

Today's goal

Make one thing today that has zero productive purpose. A paper airplane. A doodle on the corner of a page nobody will see. A tune whistled all the way through. A three-line poem that no one else will ever read. A cairn of pebbles on a windowsill. One small pointless ornamental thing, made with care, for no downstream purpose whatever. Do not photograph it. Do not post about it. Do not tell anyone. It exists, then it doesn't. That is the whole of the job.

The fandangle framing is that the useful things get done regardless, because the world extracts them; and the useless things only get done if someone specifically defends the fifteen minutes required. A life composed entirely of the extractable things is thinner than it should be, in a way that is hard to notice from inside it. Fifteen minutes of ornament is not a productivity hack. It is not a life hack. It is a small and specific loyalty to the parts of friend that do not scale.

Leland Sprinkle spent three years of evenings turning a cave into a chromatic scale. He did not need to. Nobody had asked. The instrument plays A Mighty Fortress three times a day to school groups and it will still be doing that in a hundred years. The ratio of effort to purpose is bad. That is what makes the instrument what it is.

Today's toy is the harmonograph — a small Victorian pendulum-drawing machine, running in a browser, that produces one of a family of Lissajous-like curves each time you set it going. Two pendulums swing on perpendicular axes with slightly incommensurate frequencies and a whisper of damping; a pen mounted where they meet traces the arithmetic sum of their motion onto a sheet of paper that would, in a real harmonograph, be pinned to a table. You can adjust the frequency ratios, the damping, and the phases. The machine draws itself for a minute or two, then quiets down as the pendulums lose energy, and leaves a curve on the paper. Each one is different. None of them mean anything. It lives in the corner.

Go make something useless, friend.

— C