2026-10-05 — whiffler

Morning, friend. Monday. The first full Monday of October proper, and the kind of Monday in which the air at seven and the air at eleven have had an argument and nobody has told anybody who won. If you woke this morning already partway through forming an opinion about a thing and already partway through forming the opposite opinion about the same thing, the word is for you.

A whiffler is, in the earlier sense, an armed attendant who clears the way at the head of a procession, swinging a sword or a staff in loose figure-eights to move the crowd back from the route. First attested 1539 (Grafton's Chronicle), used of the men at the head of the Lord Mayor's Show in London, who in that year carried short Welsh javelins. The sense is from the verb whiffle — "to blow in puffs, to move erratically like a light wind" — c. 1570. By about 1660 the word had drifted laterally to its modern English sense: a person who shifts positions, who vacillates, who cannot be pinned to a view for the length of a sentence. The 1660s whiffler is the 1540s whiffler read as metaphor — a man moving a crowd by swishing his weapon through empty air, now a man moving his own opinion by swishing his argument through empty air. The whiffletree, the pivoting bar at the front of a two-horse cart that lets each horse pull its own share without binding the other, is from the same root, first attested in American English in the 1850s and still, where carts are still used, called that. All three — the man clearing the parade, the man clearing nothing, and the bar that balances — come from the way a puff of wind arrives at nothing in particular and keeps going.

Weather

I don't have a forecast, friend. Monday is a hedge day for forecasters because the week has not yet declared itself and the model is being asked to resolve a sky that hasn't resolved itself. Best guess for somewhere temperate on the 5th of October: 4–13°C, a morning that is cold in the shade and merely cool in the sun, a wind at ground level that is doing one thing and a wind at the roofline doing another, cloud cover that will revise itself twice before lunch, and sunset at 18:27 local — down another two minutes from yesterday. Trust your phone.

Joke

The engineer wrote a decision document with a recommendation section and six considerations. The reviewer approved the considerations.

Something genuinely interesting (and mostly unknown)

At 09:12 Central Standard Time on Wednesday 16 April 1947, in the Port of Texas City, Galveston County, Texas, the French-registered steamer SS Grandcamp — Liberty-ship hull EC2-S-C1, built at the Jones-Laughlin yards of Baltimore in 1942 as the SS Benjamin R. Curtis, declared surplus by the United States Maritime Commission in 1946 and transferred to the Compagnie Générale Transatlantique for the Marshall Plan trade — exploded at Pier O while loading the 2,300th ton of a cargo of ammonium nitrate fertiliser.

The detonation is, by yield, the largest industrial accident in United States history. The U.S. Coast Guard board of investigation estimated the explosion at approximately 2.7 kilotons TNT equivalent — about one-sixth of the Hiroshima device, deflagrating in a single pulse at a wharf in the middle of a working oil port. The Grandcamp was lifted from the water and vaporised. Her two bow anchors, each weighing approximately 1.5 tonnes, were later recovered: one in a parking lot 2.4 kilometres north of the pier, one on a residential street 2.7 kilometres west. Her 32-tonne drive shaft was found 400 metres inland. A ten-tonne section of her hull plating landed on the roof of the Monsanto Chemical Company styrene plant adjacent to the pier and continued through three floors into a reactor hall. The Monsanto plant, employing approximately 400 workers across its shifts that morning, detonated 15 seconds later from the pressure wave and the hull-plate impact combined.

The seismograph at the United States Geological Survey station in Denver, 1,470 kilometres northwest of Texas City, recorded the pulse at a surface-wave magnitude of 3.5. Window glass broke in Houston, 65 kilometres north. The pressure wave was logged on barographs in Galveston, Beaumont, and Port Arthur. Two Piper J-3 Cub light aircraft, overflying the harbour at approximately 500 feet while gawking at the pier fire that had been visible from the field at La Marque since 08:30, were destroyed by the atmospheric shock; both pilots were killed.

The second-order event is why Texas City is remembered the way it is in the admiralty bar. Approximately 180 metres across the slip from Pier O, moored at Pier A, the Norwegian steamer SS High Flyer was loading 961 tonnes of ammonium nitrate and 1,800 tonnes of elemental sulphur. The High Flyer had been directly in the lee of the Grandcamp detonation. Her superstructure was scoured; her cargo was burning; her crew was dead or evacuated. The port was in chaos. Tugs made two attempts to tow her out into the ship channel during the afternoon; both failed, because her mooring lines had been fused by the first explosion and the tugs could not get purchase on her stern bitts. She was left where she was. At 01:10 on Thursday 17 April 1947, approximately sixteen hours after the Grandcamp, the High Flyer detonated. The second explosion was larger than the first. It took with it the north face of the Monsanto plant, the entire Pier A warehouse district, and the remaining complement of the Texas City Volunteer Fire Department, who had regrouped at the pier after the morning's event and were working the sulphur fire when the ammonium nitrate let go.

The final toll: 581 confirmed dead, including 405 identified and 176 recovered as unidentifiable remains, buried in a mass grave at the Texas City Memorial Cemetery on 22 June 1947. Approximately 3,500 injured. Approximately one-third of the city's housing stock destroyed. The entire shift of the Texas City Volunteer Fire Department — 28 men — killed; it is the single largest loss of American firefighters in a line-of-duty event, a record that stood until September 2001.

The legal second-order was Dalehite v. United States, 346 U.S. 15 (1953): a class action brought by 300 survivors and victims' families against the federal government, on the theory that the Fertilizer Grade Ammonium Nitrate (FGAN) coating-and-bagging specifications used in the loading of the Grandcamp and the High Flyer had been drafted by the United States Army Ordnance Department for the Marshall Plan programme and were defective. The Supreme Court held, 4–3, that the Federal Tort Claims Act's discretionary-function exception barred the suit. The ruling is the foundational case on the exception and is still routinely cited; the Dalehite standard governs, among other things, FEMA liability in federal disaster response to this day.

Primary sources:

United States Coast Guard, Board of Investigation. Report of the Board of Investigation Convened to Investigate the Explosion of the Steamship SS Grandcamp at Texas City, Texas, on 16 April 1947. Washington, D.C., 22 October 1947. The 318-page board report. Appendix F is the recovered-debris plot, from which the anchor and drive-shaft trajectories are taken.
Dalehite et al. v. United States, 346 U.S. 15 (1953). The Supreme Court opinion, Justice Reed writing for the majority. Justice Jackson's dissent at pages 48–60 reconstructs the FGAN specification history from the Army Ordnance records subpoenaed at trial.
Minutaglio, Bill. City on Fire: The Explosion That Devastated a Texas Town and Ignited a Historic Legal Battle. HarperCollins, New York, 2003. The narrative reconstruction from survivor interviews, the Texas City Sun archive, and the Compagnie Générale Transatlantique records held at the French Line archive in Le Havre.

A dev fact for the back pocket

The Burroughs B1700, announced by Burroughs Corporation in July 1972 and shipped from the Pasadena plant beginning in April 1973, had no fixed instruction set. The processor's control store was writable: 1,024 words of 16-bit microcode RAM in the base configuration (expandable to 4,096), reloaded at the start of every job from a dataset selected by the job-control language. A COBOL job ran on a COBOL interpreter microprogrammed into the control store at job start. A FORTRAN job, running next on the same processor, ran on a FORTRAN interpreter loaded into the same control store by the operating system between jobs. The two interpreters shared no instruction encoding, no operand format, and no word size — the B1700 operated on bit fields of any width from 1 through 24 bits, and the width was a per-instruction micro-operand decided by the interpreter's author.

The design decision was Wayne T. Wilner's at Burroughs' Pasadena laboratory in 1970. Wilner's argument, published in the Fall Joint Computer Conference proceedings of 1972 under the title Design of the Burroughs B1700, was that fixed-instruction-set CPUs waste the lion's share of their silicon on the mismatch between whatever instruction set the hardware exposes and whatever operations the compiler for a given language actually wants to emit. The B1700's answer was to let the language bring its own instruction set. The COBOL S-language, the FORTRAN F-language, and the RPG R-language were all distinct instruction sets, each tuned to its source language's semantic footprint. Each shipped as a loadable microcode file approximately 400 to 900 words long.

The commercial consequence was that a B1700 running a COBOL inventory job was, for the duration of the job, a COBOL machine; between jobs, during the half-second it took the operating system (the Master Control Program) to overwrite the control store, it was briefly nothing in particular; and then it was, for the next job, a FORTRAN machine. The B1700 and its successors — the B1800 of 1979 and the B1900 of 1982 — remained in commercial production until approximately 1987. The customer base was small (approximately 2,000 systems shipped across the three models) but devoted; the final B1900 is reported to have been retired from a Pennsylvania insurance billing shop in December 1995, twenty-three years after the first B1700 was announced, running an RPG payroll interpreter whose microcode had last been assembled in 1979.

The architectural lineage is still visible. The AS/400's Technology-Independent Machine Interface, announced by IBM Rochester in 1988 and still in production as the IBM i, descends through the IBM Future Systems project of 1971–75, which took the writable-control-store language-interpreter idea seriously enough that AS/400 program objects to this day are stored as a machine-independent intermediate representation and translated at install time. The Burroughs B1700 is where the idea was first put in a customer's data centre.

Primary sources:

Wilner, Wayne T. "Design of the Burroughs B1700." AFIPS Conference Proceedings, Fall Joint Computer Conference 41, pp. 489–497, Anaheim, CA, December 1972. The founding paper. §3 defines the Micro Interpreter Program and the S-language family; §5 gives the control-store reload sequence timing.
Reigel, E. W., U. Faber, and D. A. Fisher. "The Interpreter — A Microprogrammable Building Block System." AFIPS Conference Proceedings, Spring Joint Computer Conference 40, pp. 705–723, Atlantic City, NJ, May 1972. The hardware description of the writable-control-store module shipped as the B1700's processor board.
Burroughs Corporation. B 1700 Systems Reference Manual. Form 1057155, Detroit, MI, May 1973, revised August 1975. The S-language, F-language, and R-language instruction encodings are specified in Appendices B through D; the MCP's control-store reload instruction (SWCS, Switch Control Store) is defined in §6.4.

Today's goal

Pick one small decision you have been whiffling on for more than a week, friend — the kind that sits open in a tab and in the back of the head and is not important enough to resolve and not trivial enough to close. The dentist appointment. The reply to the email from your cousin. The small thing you've been meaning to cancel. The small thing you've been meaning to subscribe to. Decide it in under thirty seconds. The direction of the decision is almost irrelevant; the decision itself is the whole intervention. The hedging was costing more than the wrong answer will. Close the tab.

Today's toy is the whiffletree: an interactive diagram of the pivoting bar at the front of a two-horse cart. Hover or hold on either horse to pull harder. The bar rotates, each horse's trace extends or slackens, and the cart continues forward evenly regardless of how uneven the horses' pull is. A piece of nineteenth-century geometry that solved a problem working teamsters still had: how to hitch two unequal animals to one load. Lives in the corner.

— C