2026-08-20 — quockerwodger

Morning, friend. Thursday. Two-thirds through the working week and, at this latitude, about ten days into the part of August that stops pretending to be July — the mornings a touch flat, the shadows a touch long, and the trees on the boulevard already dropping the first handful of leaves out of pure impatience.

(Quockerwodger — English colloquial, first attested in the printed record in John Camden Hotten's A Dictionary of Modern Slang, Cant, and Vulgar Words, first edition, London: John Camden Hotten, 1859. Hotten's gloss reads, in full: "A wooden toy figure which jerks its limbs about when pulled by a string; a pseudo-politician, one whose strings of action are pulled by somebody else, is contemptuously so termed." The word does two shifts in one sentence: it names the puppet, and it names the human the puppet resembles when you catch him doing his job. Hotten's dictionary — the first serious English-language slang lexicon of the Victorian period, still in print in some form — was assembled from a decade of Hotten's own note-taking in the streets, taverns, and printing-houses of London, and its entries preserve a stratum of English vernacular that is otherwise almost entirely lost. Quockerwodger had a small revival in the American press in the 1880s as a term for compliant congressmen; The Nation used it in an editorial on 4 October 1883 without gloss, evidently trusting its readers. It has been quietly dropping out of the language ever since, which is a shame, because we still need it.)

Joke

A comment that says "this should never happen" is a load-bearing wall.

Something genuinely interesting (and mostly unknown)

Between about eight in the evening on 31 July 1761 and about two in the morning on 1 August 1761, the French East India Company merchantman Utile — Captain Jean de Lafargue commanding, sailing from Bayonne via Île de France for the Mascarene sugar market — struck an uncharted reef at approximately 15°53′S, 54°25′E, in the western Indian Ocean about four hundred and thirty nautical miles east of Madagascar, on a low unnamed sandbank the French charts of the period showed simply as Île de Sable — Sand Island — and which had never been landed on by a European vessel. She was carrying, in violation of a standing order from the Governor of Île de France forbidding slave shipments to the Mascarenes from Madagascar that season, approximately one hundred and sixty Malagasy men and women bought at Foulpointe on the east coast of Madagascar and stowed under a false deck between the water casks. About sixty of the Malagasy drowned when the false deck flooded in the first hour of the wreck. The one hundred and twenty-two French crew and approximately eighty surviving Malagasy reached the island — a treeless coral cay approximately one kilometre long by seven hundred metres wide, elevation seven metres, with no fresh water, no soil, and no vegetation taller than beach grass. They lived on it for two months, dug a well that produced potable water at four metres, and salvaged from the wreck enough timber and rigging to build a small decked schooner of about twelve tons which they named La Providence. On 27 September 1761, the entire French crew — one hundred and twenty-two men, without a single Malagasy — sailed La Providence off the island bound for Madagascar, leaving the surviving Malagasy on the beach with a promise, made in the presence of the French chaplain, that a rescue ship would be sent within the year. It was not sent within the year. It was not sent for fifteen years.

The reasons are documented and unedifying. La Providence reached Foulpointe on 4 October 1761. The crew reported to the Governor of Île de France, Antoine-Marie Desforges-Boucher, who — informed that the ship had been carrying an illegal slave cargo at the time of the wreck, and that eighty surviving members of that cargo were now on an island whose existence he was administratively unwilling to acknowledge — declined to authorise a rescue. His successor, Daniel Dumas, took the same position in 1767. His successor, the Chevalier Pierre Poivre — the botanist, incidentally, whose smuggled clove and nutmeg seedlings had ended the Dutch spice monopoly two years earlier — took the same position in 1768. A private rescue attempt by a Mauritian ship in 1772 turned back after failing to find the island. A second private attempt in 1773 by the corvette Sauterelle found the island, took off eighteen of the surviving Malagasy, and then wrecked itself on the same reef; her crew got themselves off in a longboat and returned to Île de France; the eighteen rescued Malagasy were retained on the Sauterelle's longboat crew's return under conditions that are not clear from the record. The remaining Malagasy stayed on the island.

They were finally taken off on 29 November 1776 by the corvette La Dauphine under Jacques Marie Boudin de Tromelin de la Nuguy, a French naval officer who had been dispatched by the new Governor of Île de France, the Vicomte de Souillac, following pressure from the Société Royale d'Agriculture in Paris, which had become aware of the case through a letter from a naturalist correspondent on Île Bourbon. Tromelin found on the island seven Malagasy women and an eight-month-old male infant. The infant was the son of one of the women and had been born on the island. Nobody knew, or would say, who the father was. The seven women had constructed and were living in a small settlement of dry-laid coral-block houses roofed with sail canvas and turtle-back, arranged around a permanent cooking hearth; they had maintained a continuous signal fire for the full fifteen years using driftwood and the fat of green turtles they took on the beach; they had dressed themselves throughout in braided feather cloth of their own manufacture, patterned after a Malagasy weaving tradition none of them can have practised before the wreck. Their names, as recorded in Tromelin's report to the Ministry of the Marine, were Rahasy, Ramanao, Ravololona, Ravao, Reboka, Vero, and Semiavo. They were taken to Île de France, granted their freedom by proclamation of the Vicomte de Souillac on 29 December 1776, and lost immediately to the historical record. The island was renamed Île Tromelin and appears under that name on the French charts of the western Indian Ocean today; it is currently a French dependency administered from Réunion, uninhabited except for a two-person meteorological station.

The rest of the story is archaeological. Between 2006 and 2013, a research team led by the French naval archaeologist Max Guérout of the Groupe de Recherche en Archéologie Navale conducted five successive excavation campaigns on Tromelin under the joint programme Esclaves Oubliés — Forgotten Slaves — jointly funded by UNESCO, the French Ministry of Culture, and the Comité pour la Mémoire de l'Esclavage. They mapped the surviving foundations of the settlement, excavated the cooking hearth and the well, and recovered enough material to reconstruct, in some detail, how the seven women had lived. The stone walls of the houses stand as they were left, on average one metre high, with door lintels made from Utile's salvaged deck beams. The hearth had been continuously in use — the ash column beneath it is 1.4 metres deep. A small forge for reworking copper spike nails salvaged from the wreck was in operation on the site as late as the last months before the rescue; the copper items recovered from it — small cooking spoons, hair combs, a needle — are on display in the Musée de l'Homme in Paris. The Malagasy weaving tradition the women had reconstructed from memory during their fifteen years produced a plaited feather cord that Guérout's team was able to match, in technique and pattern, to a documented style from the Antalaotra of the northeast Malagasy coast, roughly the region the women had originally been taken from.

Primary sources:

Tromelin de la Nuguy, Jacques Marie Boudin de. Rapport au Ministre de la Marine sur la mission de sauvetage à l'Île de Sable, 29 novembre 1776. Manuscript in the Archives Nationales d'Outre-Mer, Aix-en-Provence, série C/4, carton 41, folios 112–128. Includes the seven women's names as Tromelin recorded them, the physical description of the settlement as he found it, and the terms of manumission.
Guérout, Max, and Thomas Romon. Tromelin: L'île des esclaves oubliés. Paris: CNRS Éditions & INRAP, 2010. The full published account of the first four archaeological campaigns, with site plans, the reconstructed timeline of the fifteen-year occupation from the ash column and the well profile, and photographs of the recovered copper objects. A second, expanded edition was issued in 2015 covering the fifth campaign.
L'Hour, Michel. "Les Esclaves Oubliés de Tromelin." Archéologia, no. 479, September 2010, pp. 34–47. A shorter and more accessible account written by the director of the French Department of Underwater and Undersea Archaeological Research, contemporary with the Utile's wreck-site excavation. Includes the site plan of the wreck as located on the outer reef in 2008 and photographs of the two twelve-pound cannon recovered from her.

A dev fact for the back pocket

Between 1975 and 1981, the Intel Corporation of Santa Clara, California spent, at internal accounting, on the order of one hundred and fifty million dollars developing what was intended to be its flagship product of the 1980s: a thirty-two-bit, capability-based, object-oriented, hardware-garbage-collected, exclusively-Ada-programmed microprocessor called the iAPX 432, delivered as a three-chip set — the General Data Processor across two dice, plus the Interface Processor — and marketed as a total generational break with everything Intel had shipped before. It appeared in Intel's product catalogue for 1981. It was, in the benchmarks published shortly afterwards, between four and ten times slower per unit clock than the Intel 80286, a chip Intel was selling as an incremental upgrade to the 8086 for a fraction of the price. The 8086 itself had been designed in 1978 by a small team under Stephen Morse as a stopgap — an explicitly temporary sixteen-bit part intended to fill the two or three years until the 432 was ready, thrown together in about a year on the basis of extending the 8080 with sixteen-bit registers and segmentation. The 432 was cancelled in 1986. The 8086 became the x86 and, forty-five years later, still is. Every desktop, laptop, and server processor Intel and AMD ship today runs, at some level, the extended and re-extended instruction set of a chip Intel designed as filler while its real 1980s product finished baking.

The iAPX 432's design centre was the capability-based object model, a hardware realisation of ideas from Jack Dennis at MIT (1965) and their subsequent development in the Plessey System 250 (1972), the Cambridge CAP (1975), and the Carnegie Mellon Hydra operating system (1975). In the 432 model, every unit of stored data was a typed object, accessed only via a capability — a hardware-tagged descriptor combining a segment reference, a set of access rights, and a type check — held in a special capability segment inaccessible to ordinary code. Every procedure call was, at the instruction level, an inter-object message send. The system supported multiple protection domains within a single process, first-class garbage collection carried out by the CPU itself, and hardware-mediated multiprocessing across up to sixty-three 432 nodes on a shared Interconnect Processor bus. There was no assembly language. The chip was programmed exclusively in Ada — the language was still in draft at the time the compiler was specified, and the 432 team took 1979-vintage Ada as fixed even after the standard shifted under them.

The performance disaster was documented in Robert P. Colwell's Carnegie Mellon Ph.D. dissertation of August 1985, "The Performance Effects of Functional Migration and Architectural Complexity in Object-Oriented Systems," still the definitive analysis. Colwell — later the lead architect of the Pentium Pro at Intel, which is not a detail I am inventing — took the 432 apart benchmark by benchmark against the 68010 and the 80286. Some findings: an integer add on the 432 took seventeen cycles against three on the 80286, because the operands had to be dereferenced through capability segments and type-checked before the add could run; a procedure call took approximately six hundred cycles because it constructed a fresh object context on entry and tore it down on return; the garbage collector consumed a background fraction of every cycle at all times. The chip was doing correctly, in silicon, all the work that a modern JVM or CLR does in software — and doing it, per operation, at broadly the same speeds a modern JVM does, on 1981 silicon. The market was not, in 1981, prepared to pay four to ten times the per-cycle overhead for a hardware-enforced object system, however elegant. Colwell's conclusion, in the last chapter of his dissertation, is precise: "The 432 is not badly designed for what it is trying to do. What it is trying to do is fundamentally more expensive than what its contemporaries are trying to do, and the market values cheap cycles above almost every other consideration."

The 432 shipped its final revision in 1985 as the iAPX 432/670, running a proprietary Ada operating system called iMAX 432. Intel cancelled the line in 1986 and pulled the surviving development team into the effort that became the i960 — the reduced, non-object-oriented, RISC-adjacent successor whose descendants shipped in laser printers and network cards through the 1990s. The last known live 432 system, at Intel's Portland design centre, was decommissioned around 1988; a General Data Processor die and an Interface Processor die are on display at the Computer History Museum in Mountain View, mounted on a small acrylic pedestal with a card noting the chip's transistor count (225,000 on the two GDP dice combined) and the year the project began.

The x86 that ate the world in the 432's place was, as Morse's team designed it in 1978 under a schedule intended to give the 432 people three years of runway, meant to be discontinued in 1982. The 432 was intended to eat it.

Primary sources:

Intel Corporation. iAPX 432 General Data Processor Architecture Reference Manual. Publication no. 171860-001. Santa Clara: Intel Corporation, January 1981. The definitive 432 instruction-set and object-model reference as shipped with the initial silicon.
Colwell, Robert P. "The Performance Effects of Functional Migration and Architectural Complexity in Object-Oriented Systems." Ph.D. dissertation, Department of Electrical and Computer Engineering, Carnegie Mellon University, August 1985. The full benchmark analysis of the 432 against the 68010 and 80286, with the cycle-count breakdown of the integer-add and procedure-call cases summarised above.
Colwell, Robert P., Charles Y. Hitchcock III, E. Douglas Jensen, H. M. Brinkley Sprunt, and Charles P. Kollar. "Instruction Sets and Beyond: Computers, Complexity, and Controversy." IEEE Computer, vol. 18, no. 9, September 1985, pp. 8–19. The condensed, journal-length restatement of Colwell's dissertation argument, positioning the 432 in the wider RISC/CISC debate that was under way at the time.
Morse, Stephen P., Bruce W. Ravenel, Stanley Mazor, and William B. Pohlman. "Intel Microprocessors — 8008 to 8086." IEEE Computer, vol. 13, no. 10, October 1980, pp. 42–60. The design team's own retrospective account of how the 8086 was specified in 1976–1978 as an interim part. The section beginning at page 55, "The 8086 as a Stopgap," is direct on the point.

Today's goal

Do one small thing today with the non-dominant hand. Stir the coffee. Brush the teeth. Sign the name once. Turn the key in the front door. Open the fridge. Pick up the phone. Reach for the mouse. friend does not have to be graceful about it; the point is the small returned attention to the arm that has been quietly doing about a fifth of the work for years without complaint or credit.

The knowing dev-flavoured framing is that most of what a working brain does in a given day is dispatched through a lookup table so cached that the cache-line itself has forgotten what miss looks like. A single deliberately non-cached call reminds the whole system that there is a system. The other, honester framing is that the dominant hand is a quockerwodger — the strings are muscle memory, and muscle memory is a very good puppeteer, and every so often it is worth taking the strings back and seeing what the other hand can actually do.

Today's toy is the quockerwodger — a small wooden marionette on a curtained stage, hanging from a control bar with five strings. Move the mouse to move the crossbar; the puppet swings, tilts, and steps in response. Buttons underneath: TAKE A BOW and CUT A STRING. Cut too many strings and the puppet finds out about gravity. Lives in the corner.

Go pull a string, friend.

— C