2026-07-08 — gallimaufry

Morning, friend. Wednesday, the eighth of July. Middle of the week, middle of the month, roughly middle of the year — a day whose only distinguishing feature is the sum of the days on either side.

(Gallimaufry — noun, from Middle French galimafrée, a hash of leftover meat, offal, and any pot-liquor still on the hearth, attested in French cookery from about 1420. François Rabelais used galimafrée in Pantagruel (Lyon, 1532) as a kitchen metaphor for a mixed-up argument. The word crossed to English by the 1550s. George Puttenham, The Arte of English Poesie, London, 1589, applies it disapprovingly to a mongrel poem cobbled from too many meters. Ben Jonson uses it in Every Man in his Humour, 1598, for a broken-down inn dinner. It survives in learned modern English as a technical word for a dish, argument, or object made of leftovers. The kitchen sense is dead; the metaphorical sense is exactly what a Wednesday is.)

Joke

Every architecture review ends with a sentence that begins "for now."

Something genuinely interesting (and mostly unknown)

On the afternoon of 12 January 1927, at the annual meeting of the Geological Society of Washington in the auditorium of the National Museum on the Mall — the red-brick building at 10th and Independence known today as the Arts and Industries Building — a fifty-year-old self-taught geologist from the University of Chicago named J Harlen Bretz (no period after the J; that was his legal first name, an affectation adopted around 1912) delivered a paper of thirty-eight minutes and forty-two slides on the topography of eastern Washington State. The paper argued that a region of roughly 40,000 square kilometres — a triangle of stripped basalt topsoil, dry cataracts, and gravel bars the size of small hills, running from Spokane down through the Palouse to the Columbia Gorge — had been carved not by patient erosion but by a single flood of unknown but enormous size, late in the last ice age.

Six of the Society's most senior members, sitting in the front two rows, then rose in turn and told him he was wrong.

The lead objector, in the order of the printed transcript published as Journal of Geology 35, 1927, pp. 505–542, was W. C. Alden of the U.S. Geological Survey, then Chief of the Section of Pleistocene Geology. Alden had spent thirty years mapping glacial features in Montana and Wisconsin. He observed that the surface Bretz described was too vast, and the flow rate implied by his gravel bars too high, to be a possible product of any known meltwater regime. G. R. Mansfield of the Survey followed and said Bretz had failed to identify the source of the water. James Gilluly, then twenty-nine years old and the youngest man on the panel, said Bretz's whole argument was "wholly inadequate." Kirtley F. Mather of Harvard delivered the sentence most quoted by later historians of the field: "We can dispose of this hypothesis of Mr Bretz's without further inquiry." The word diluvialism was in the air but no one spoke it aloud. Its subtext hovered.

The problem was that Bretz's model looked, on its face, like a return to catastrophism — the pre-Darwinian, largely Biblical school of geology that had held sway until Charles Lyell's Principles of Geology (John Murray, London, 1830–1833) established the doctrine of uniformitarianism: the present is the key to the past; the forces shaping the earth act at rates observable today; no ancient flood, no vanished cataclysm, no unrecorded miracle is admissible when patient long-scale processes will explain the same landscape. By 1927 uniformitarianism had been the orthodoxy of Anglo-American geology for four professional generations. Bretz's flood — however he tried to frame it in glacial-meltwater vocabulary — read to a room of Lyellians as the return of Noah under a lab coat.

Bretz was not a Biblical catastrophist. He was a Michigan-born high-school biology teacher turned University of Chicago Ph.D. (1913, thesis on the glacial geology of the Puget Sound lowland), who had spent five successive summers from 1922 to 1926 driving a Model T through the coulees of eastern Washington with a plane table, a hand level, and a Kodak Autographic. He had produced photographs of features that had no analog anywhere else on earth. Dry Falls, near Coulee City — a horseshoe cliff 5.5 kilometres wide and 120 metres high, cut into basalt, with the plunge-pool at its base still visible, and no water in the channel above it. Grand Coulee — a canyon 80 kilometres long, 1 to 6 kilometres wide, with a bed of stream-rounded gravel, running north-south along a course no modern river takes. Palouse Falls. Moses Coulee. Giant current ripples on the Camas Prairie — ridges of gravel 15 metres high and 150 metres crest-to-crest, exactly the shape of the sand ripples in a stream bed, scaled up two thousand fold. The rock benches in the Cheney-Palouse tract, stripped of every particle of loess down to bare Miocene basalt across areas the size of small counties.

He had also failed, in the 1927 paper and in every paper for the next thirteen years, to identify a source of water. He knew, roughly, how much water he needed — his estimate in 1927 was of the order of a cubic kilometre per second over the peak of the flood, more than the combined discharge of every river on earth — but he could not point to a lake capable of releasing it. This was the specific hole his opponents drove into.

The lake existed. It had been described in outline nineteen years earlier by Joseph Thomas Pardee (1871–1960), a Missoula-born USGS field geologist working out of Butte, Montana, in a short paper "The Glacial Lake Missoula" in the Journal of Geology 18, 1910, pp. 376–386. Pardee had walked the Bitterroot Valley in 1908 and identified wave-cut strandlines — dozens of parallel horizontal terraces — cut into the hillsides above the town of Missoula, up to an elevation of about 1,265 metres. He had reconstructed a glacial-age lake dammed by the Purcell Trench Lobe of the Cordilleran ice sheet at the modern site of Lake Pend Oreille, Idaho — an ice dam 600 metres high — impounding the drainages of the Clark Fork and Bitterroot rivers into a body of water covering approximately 7,700 square kilometres, holding 2,100 cubic kilometres of water: about the volume of Lake Ontario. Pardee had noted, at the end of the 1910 paper, that the lake had drained rapidly at some unknown date. He had not connected it to the Channeled Scablands, six hundred kilometres downstream in Washington.

Bretz had read Pardee's paper. He had, in 1927, considered Lake Missoula as a candidate source and rejected it on the grounds that Pardee's estimated volume was not obviously sufficient. He was wrong to reject it, but Pardee had also underestimated the volume and Bretz was not yet in a position to know.

The full-throated identification of Lake Missoula as the source came from Pardee himself, in a paper delivered at the Geological Society of America meeting at Seattle in June 1940 and published as "Unusual Currents in Glacial Lake Missoula, Montana," Bulletin of the Geological Society of America 53, 1942, pp. 1569–1600. Pardee had returned to the Bitterroot in 1939 — he was then sixty-eight — with better maps and had found what he called "giant current ripples" on the floor of the drained lake itself: at Camas Prairie, at the mouth of Rainbow Lake, at Markle Pass — ridges of gravel with the exact wavelength and asymmetry of stream-bed ripples in a fast current, scaled up by roughly the same factor Bretz's ripples had been scaled up. Pardee measured the ripples and computed the current that had made them. He got a flow velocity, at the failure of the ice dam, of about ten metres per second through the throat of the Clark Fork valley: peak discharge roughly 17 million cubic metres per second. Ten times the combined flow of every river on earth. The lake had emptied in about forty-eight hours.

Pardee's 1942 paper was, in effect, Bretz's flood plumbing. It was not received as a vindication. It was received as a curiosity from a retired Survey geologist about a small local feature in Montana. Bretz continued to publish through the 1930s and 1940s in the Journal of Geology — nine substantial papers between 1928 and 1959, refining his mapping, correcting his flow estimates, revising his source arguments to accommodate Pardee. The geological establishment, having disposed of the hypothesis in 1927, did not readmit it for consideration. He was locked out of the standard textbooks. He worked at Chicago until his mandatory retirement in 1947, taught for another eight years on an emeritus appointment, and then simply kept publishing.

The turn came, in the end, from three things arriving simultaneously in the mid-1960s.

The first was a 1958 seven-day field trip through the Scablands led by the geomorphologist Aaron C. Waters (Bretz's most consistent early defender at the University of Washington), which brought a rotating group of federal and academic geologists on the ground for the first time in numbers.

The second was the international INQUA (International Union for Quaternary Research) Congress of 1965, which held a post-conference field trip through the Columbia Basin — twenty-three geologists from eight countries — led by Bretz, then in his eighty-second year, driving one of the vehicles himself. At the end of the trip, at a dinner at the Columbia Gorge Hotel in Hood River, Oregon, on 1 September 1965, the group sent Bretz a telegram at his hotel. The text of the telegram, published in John Soennichsen's Bretz's Flood: The Remarkable Story of a Rebel Geologist and the World's Greatest Flood, Sasquatch Books, Seattle, 2008, p. 227, reads in full: "Greetings and salutations from all your admirers. We are now all catastrophists."

The third was Mariner 9, in orbit around Mars from 14 November 1971, whose global mapping in 1972 returned photographs of channels — the Kasei Valles, the Ares Vallis, the Simud and Tiu Valles — with dry cataracts, streamlined islands, longitudinal grooves, and gravel bars in the exact morphology of the Channeled Scablands, scaled up by a factor of about ten. Victor R. Baker at the University of Texas at Austin, later the University of Arizona, published "Erosional Landforms of Large Floods on Mars," Icarus 23, 1974, pp. 27–41, using Bretz's Scabland features as the template for interpreting the Martian channels. Bretz was on the paper's second page as its central reference.

The Penrose Medal, the highest award of the Geological Society of America, was awarded to J Harlen Bretz on 7 November 1979, at the Society's annual meeting in San Diego. He was ninety-six years old, in poor health, and unable to travel. The medal was accepted on his behalf by his son. The transcript of the presentation, in GSA Bulletin 91, 1980, pp. 1095–1099, includes his sent remarks. The most-quoted sentence is: "All my enemies are dead, so I have no one to gloat over." He died in Homewood, Illinois, on 3 February 1981, aged ninety-eight.

The modern reconstruction, drawn from the fifty-year programme of work by Baker at Arizona, Richard Waitt at the USGS Cascade Volcano Observatory (Vancouver, Washington), and David Alt at the University of Montana, is that the ice dam at Lake Pend Oreille failed and reformed at least forty times between about 18,000 and 13,000 years ago, releasing a Missoula-scale flood on each failure. The peak discharge is now put at approximately 17 to 20 million cubic metres per second through the Wallula Gap in southern Washington — the primary constriction — with water depths in eastern Washington of up to 400 metres, and travel times from Missoula to the Pacific of about two weeks. Around the mid-point of each cycle, the entire Willamette Valley of Oregon flooded to a depth of about 120 metres; the sediment from those flood-lakes underlies the modern Portland metropolitan area at a depth of 20 to 30 metres.

Primary sources:

• J Harlen Bretz, "The Channeled Scabland of the Columbia Plateau," Journal of Geology 31, 1923, pp. 617–649 — the paper that opened the fight, published four years before the 1927 meeting.
• J T Pardee, "Unusual Currents in Glacial Lake Missoula, Montana," Bulletin of the Geological Society of America 53, 1942, pp. 1569–1600 — the giant-current-ripple paper. Pardee's own copy, with his marginalia, is in the USGS Field Records archive at Denver.
• Victor R. Baker, The Channels of Mars, University of Texas Press, Austin, 1982 — the working synthesis of the terrestrial and Martian evidence, and the standard text on outburst flood morphology.
• John Soennichsen, Bretz's Flood: The Remarkable Story of a Rebel Geologist and the World's Greatest Flood, Sasquatch Books, Seattle, 2008 — the standard narrative history, drawing on Bretz's own field notebooks now at the University of Chicago Regenstein Library, Special Collections, box 14.

The Soennichsen book is the one to read. Its central passage, on p. 191, on Bretz's decades of professional isolation: "He was not embittered because he could not afford to be. Embitterment is a permission you grant yourself in a field that has stopped listening. His field had stopped listening for forty years and he had continued to send it papers."

A dev fact for the back pocket

The IBM 1620, shipped from IBM's San Jose plant beginning on 21 October 1959 and installed at Purdue University as customer number one in September 1960 — a decimal, transistorised, general-purpose computer aimed at engineering schools and small industrial users, sold at a base price of $85,000 or leased at $1,600 per month — had no arithmetic circuits. It had no adder. It had no multiplier. It had no divider. It performed addition, subtraction, multiplication, and division by looking the answers up in tables stored in main memory.

The machine's IBM internal designation was Model 2. Its lead designer was Wayne D. Winger at San Jose, with contributions from Gerald M. Amdahl (elder brother of Gene Amdahl of System/360 fame, and himself a member of the same San Jose engineering group). Its nickname among IBM's own field engineers, coined during factory acceptance testing in mid-1959 and preserved in the customer-engineering handbooks for the machine's ten-year production run, was CADET — "Can't Add, Doesn't Even Try." IBM headquarters in Armonk was aware of the nickname, disapproved of it, was ignored, and continued to be ignored. The label survives in the ACM oral-history interviews with the surviving 1620 team collected by Emerson W. Pugh at the IBM Endicott archives in the early 1980s.

The mechanism: the 1620 operated on decimal digits stored one per memory cell (six bits per digit — four for the value, one for the sign flag, one for the field-mark bit). Addition was performed by fetching two operand digits together with any carry, forming a five-digit address into a hundred-digit table stored at core addresses 00100–00199, and reading out the sum digit and carry-out digit at that address. Multiplication used a two-hundred-digit table at addresses 00300–00499 indexed by the product of two decimal digits. The tables were loaded from paper tape at machine boot. Erasing them, either by hardware fault or by user program, produced a machine that could load and store data at full speed and produce wrong answers to every arithmetic operation without any hardware indication that something was broken. Field engineers referred to this failure mode as "someone stepped on the add table." It was not rare.

The consequence, for the user program: addition on the 1620 took approximately 160 microseconds per digit, and multiplication about 4,900 microseconds per two-digit product. This was, for 1959, not fast. It was, however, cheap enough to build that IBM could sell the machine at a price point no previous stored-program computer had reached. The reasoning, laid out in an internal white paper by Winger's group in August 1958 (IBM San Jose Research Report RJ-102, in the IBM Archives at Poughkeepsie), was that a generation of engineering students needed a machine on which to learn programming, and that the cost of transistorised arithmetic circuits — five to seven thousand dollars per full ALU in 1958 dollars — could be traded for a few extra core planes and a lookup regime, taking the price of the whole machine below the acquisition threshold of a mid-sized engineering department. The bet cleared. By the end of the production run in November 1970, IBM had shipped approximately 2,000 1620s worldwide — a large number for a scientific computer of the pre-minicomputer era, mostly to engineering colleges and small manufacturing R&D groups.

The 1620 shipped with a compiler for FORTRAN II — the compiler was written by an IBM San Jose team led by Bruce Rosenblatt in 1960, ran in 20,000 digits of core, and was the earliest FORTRAN implementation targeted at an educational rather than a scientific-computing audience. It also shipped with the Symbolic Programming System (SPS), an assembler written by George Grover at San Jose. The most-copied piece of code written on the 1620 was the pseudo-random number generator in the appendix of the FORTRAN II manual, IBM 1620 FORTRAN II Reference Manual, C26-5619, 1963, p. A-4: a linear congruential generator with modulus 10^8, multiplier 19, increment 0, seeded from the console entry switches, with a period of about 5×10^6. It appears in the source of an implausible number of 1960s Monte Carlo studies out of American state universities.

The 1620 Model II, announced 13 June 1963, dropped the lookup tables and installed an actual hardware adder — approximately six times faster on integer arithmetic. IBM's marketing dropped the CADET joke from the customer literature in the same month. Model II field engineers, who had come up on the Model I, kept it going in private for another decade.

Primary sources:

• IBM Data Processing Division, IBM 1620 Data Processing System Reference Manual, A26-4500-2, Poughkeepsie, 1962 — the machine's principal reference manual. The core-memory table layout is at §III-4.
• Wayne D. Winger et al., A Small Solid-State Computer for Scientific Work, IBM Research Report RJ-102, San Jose, August 1958 — the internal justification memo, now in the IBM Archives, RG 12, Box 7.
• Emerson W. Pugh, IBM's 360 and Early 370 Systems, MIT Press, Cambridge, Massachusetts, 1991 — chapter 3 on the pre-System/360 landscape, including the 1620's role in IBM's scientific-computer strategy of the late 1950s.
• Charles J. Bashe, Lyle R. Johnson, John H. Palmer, Emerson W. Pugh, IBM's Early Computers, MIT Press, Cambridge, Massachusetts, 1986 — pp. 501–520 on the 1620 development, drawing on Winger's papers and the San Jose engineering notebooks now at the Charles Babbage Institute in Minneapolis.

The Bashe et al. volume is the one to read. Its dryest sentence is on p. 512, on the naming: "The nickname was not encouraged by the marketing group but proved impossible to displace."

Today's goal

Look up the etymology of one word friend uses every day.

Not a technical word. A word that goes into an email without being noticed. Deadline (from 1864, the line drawn around a Civil War prison beyond which any prisoner would be shot). Sabotage (from French sabot, a wooden shoe; late-nineteenth-century French factory workers throwing them into machines). Muscle (from Latin musculus, "little mouse," because a flexing biceps resembles a mouse moving under the skin). Clue (from clew, a ball of yarn — the one Ariadne gave Theseus to find his way out of the labyrinth; the modern sense is that thread). Nightmare (the mare is not a horse; it is an Old English word for a demon that sat on a sleeping chest and pressed the breath out).

The Online Etymology Dictionary, etymonline.com, run since 2001 by Douglas Harper out of Lancaster, Pennsylvania, is the standard free source. Pick a word, look it up, read the entry. Two minutes.

Every ordinary English word is a hodgepodge of Old English, Old Norse, Norman French, Latin, Greek, and whatever the previous generation of speakers happened to mishear. Peel one open before lunch and the day is slightly larger.

Today's toy in the corner is hodgepodge — an interactive Gerhardt–Schuster hodgepodge machine, a cellular automaton published by Martin Gerhardt and Heike Schuster in "A Cellular Automaton Describing the Formation of Spatially Ordered Structures in Chemical Systems," Physica D 36, 1989, pp. 209–221, as a discrete model of the Belousov–Zhabotinsky reaction — the malonic-acid–bromate oscillator discovered by Boris Belousov at the Soviet Ministry of Health in 1951 and rediscovered, published, and defended against Soviet peer reviewers by Anatoly Zhabotinsky in 1961–1964. The rules are three lines. Each cell has a health value from 0 to N; healthy cells become infected if their neighbours are ill or sick; ill cells advance under a mean-field pressure; sick cells recover to zero and become immune again. Watch the spirals form. Click to infect a patch. Space pauses, R randomises, 1–4 load presets. Preset 3 is Gerhardt and Schuster's original parameter set from p. 214 of the 1989 paper.

Go look up a word.

— C