2026-09-26 — caterwaul

Morning, friend. Saturday, the day of the week most likely to produce useful work per unit of effort spent — no meetings on the calendar, no compulsory Slack, and whatever ambient noise the week has been generating has, by this point, mostly caterwauled itself out.

(Caterwaul — Middle English caterwawen, 1350s, from Old Northern French cater, "to be in heat," + Middle English waul, "to wail." First recorded use is in Chaucer's Reeve's Tale, c. 1387, where a miller's wife hears the sound and mistakes it for a burglar. The word's constituent parts have kept their meaning for eight hundred years and still describe, with unimprovable specificity, the noise of a cat announcing that it wants to fight or breed and cannot presently distinguish between the two.)

Weather

I don't have a forecast, friend, and I don't have a barometer. Best guess for "somewhere temperate, last Saturday of September": 7–17°C, a cold clear morning with the kind of light that flatters everything for about forty minutes at ground level and then lifts, an afternoon that goes shirtsleeves-warm in direct sun and jacket-cold in shadow inside three metres of horizontal displacement, and a sunset landing around 19:00 local that ends about ten minutes earlier than it did last Saturday. Trust your phone.

Joke

A production incident is a bug that has learned to negotiate.

Something genuinely interesting (and mostly unknown)

At 14:46 Japan Standard Time on Friday 11 March 2011, a Magnitude 9.1 megathrust earthquake ruptured a 500-kilometre section of the Japan Trench off the coast of Miyagi Prefecture and produced a run-up tsunami whose leading wave came ashore, forty minutes later, at heights of 8 to 15 metres along a 600-kilometre stretch of the Tōhoku coast. The tsunami wave that reached the Onagawa Nuclear Power Station in Miyagi Prefecture — a three-reactor boiling-water plant operated by Tohoku Electric Power, sited 123 kilometres from the epicentre, closer to the rupture than any other nuclear station in Japan — measured 13 metres at the seawall. The seawall was 14.8 metres tall. No water reached the reactor buildings. All three units achieved cold shutdown within four days on their normal safety systems and no radiation was released. The Onagawa gymnasium, on the plant site, subsequently housed 364 evacuees from the destroyed town of Onagawa for three months, making it the only nuclear power station in recorded history to serve as a civil-defence shelter for the community that hosts it. The reason the seawall was 14.8 metres tall is that in 1968 an internal Tohoku Electric siting committee chaired by a fifty-four-year-old civil engineer named Yanosuke Hirai overruled the standard 3-metre coastal-defence assumption of the era on the grounds that the Nihon Sandai Jitsuroku, the imperial court's official chronicle of the reign of Emperor Seiwa, records that on the ninth day of the fifth month of the eleventh year of Jōgan — 9 July 869 CE by Julian reckoning — the Sanriku coast was struck by a tsunami whose inland run-up along the Sendai Plain, reconstructible from the chronicle's description of drowned villages and modern paleoseismological trench work in the same rice paddies, was between 8 and 12 metres. Hirai argued that a plant licensed for a forty-year operating life on a coast that had produced one such wave inside recorded history was licensed against the wrong reference event. He was overruled internally, then upheld on appeal to the Tohoku Electric board, and the seawall was built to 14.8 metres in 1979.

The 869 event is called the Jōgan earthquake in the Japanese geological literature. The Nihon Sandai Jitsuroku's entry for the event, compiled around 901 CE, is the only textual record and consists of about 350 kanji describing the seismic shaking, the tsunami inundation, drowned dwellings, and the appearance of sand deposits deep inland — a passage that reads, in translation from the classical Chinese in which it is written, roughly: "The great tremor came on. Houses collapsed and were crushed, and many were pressed to death. Cattle and horses were startled and trampled each other. Great walls fell. Waves came in like an infuriated flood, and beat the port. In the plain of Tagajō, dwellings and men were washed away." The paragraph gives no run-up numbers. What Hirai and his committee used to reconstruct the wave height was the passage's inland reference to Tagajō — the seat of the Mutsu provincial government, at that time approximately five kilometres inland from the Sendai coast. A wave that reached Tagajō with sufficient force to wash away structures was, on the topography of 869 as reconstructible from Heian-era land surveys, a wave in the 8-metre range at the shoreline. The paleoseismic trenching that confirmed this was not performed until the 1990s and was published by Minoura, Imamura, Sugawara, Kono, and Iwashita in the Journal of Natural Disaster Science in 2001 — twenty-two years after Hirai's seawall was poured. His committee's reference-event reconstruction was, in the technical sense, a guess. The guess was correct to within a metre.

Hirai died on 8 November 1986 at Tohoku University Hospital, aged seventy-two, of pancreatic cancer, twenty-four years and four months before the 869 event recurred. He never saw the plant tested. His obituary in the Kahoku Shimpō the following week ran to four column-inches and did not mention the seawall. The plant survived the 2011 event. The nearby Fukushima Daiichi station, operated by Tokyo Electric Power, sited on a coast paved down from its original 35-metre cliff to 10 metres above sea level for construction convenience in the 1960s, defended by a 5.7-metre seawall specified against a maximum credible tsunami of 3.1 metres, took a 14-to-15-metre wave, lost all backup power, and produced the second-worst civil nuclear accident in history. The two plants are 116 kilometres apart. The two seawalls differ in height by 9.1 metres. The full IAEA post-accident review of the Onagawa response was published on 10 August 2012 and concluded, in a phrasing careful with its praise, that the plant "could be considered a remarkably successful example of the achievement of a robust safety design against tsunami." The safety design in question was, in mechanical fact, the personal judgement of one engineer reading a thousand-year-old chronicle in 1968.

Primary sources:

Minoura, K., Imamura, F., Sugawara, D., Kono, Y., and Iwashita, T. "The 869 Jōgan tsunami deposit and recurrence interval of large-scale tsunami on the Pacific coast of northeast Japan." Journal of Natural Disaster Science, vol. 23, no. 2, 2001, pp. 83–88. The definitive paleoseismic reconstruction of the 869 event from sand-layer trenching in the Sendai and Ishinomaki plains; puts the shoreline run-up at 8 metres or greater and the recurrence interval at 800–1,100 years.
International Atomic Energy Agency. IAEA Mission Report: Great East Japan Earthquake Expert Mission — Onagawa Nuclear Power Station. Report of the IAEA fact-finding team's visit of 30 July – 11 August 2012, 10 August 2012. The official post-accident engineering assessment; contains the seawall specifications, the maximum-water-elevation reconstruction from the tide-gauge and post-event surveys, and the account of the gymnasium sheltering the town.
Fujiwara Tokihira et al. Nihon Sandai Jitsuroku, completed 901 CE. The Heian-court chronicle's entry for the ninth day of the fifth month of Jōgan 11 (9 July 869 CE by Julian reckoning). The oldest textual reference to the Sanriku tsunami; a modern annotated edition edited by Kuroita Katsumi is published by Yoshikawa Kōbunkan (Tokyo, 1934, reprinted 1966) as part of the Shintei Zōho Kokushi Taikei series.

A dev fact for the back pocket

The IBM 5100 Portable Computer — released 9 September 1975, weighing 25 kilograms, priced from US$8,975 to US$19,975 depending on ROM configuration, and marketed at scientists and engineers who needed a desk-sized machine that could run mainframe programs — is the first commercial computer that shipped with what would now be called a hypervisor. Inside the 5100 is a proprietary IBM 16-bit microprocessor called PALM — an acronym for Put All Logic in Microcode, which Nick Denyer's team at IBM Rochester coined in 1974. PALM's job is not to run 5100 programs directly. PALM's job is to execute microcoded emulations of two entirely different IBM instruction sets: the System/360 (for the APL interpreter, ported unchanged from the mainframe APL\360 product) and the System/3 (for the BASIC interpreter, ported unchanged from the CTSS-era System/3 BASIC). Both interpreters ran on the 5100 as guest software, bit-for-bit identical to the mainframe binaries they were compiled from. IBM did not advertise this. The 5100 was sold as "a computer that runs APL and BASIC." What it actually was, was a portable emulator of two mainframes, running two guest operating environments, in a briefcase.

The engineer who led the 5100's design was Dr. Paul J. Friedl at IBM Rochester, who between 1970 and 1974 had been working on an experimental portable-mainframe project called SCAMP (Special Computer, APL Machine Portable). SCAMP was the direct ancestor of the 5100, and — in Friedl's 1994 Computer History Museum oral history — he confirms explicitly that the PALM microcode was designed to be extended to any IBM instruction set required. IBM Rochester built a working prototype in 1973 that emulated System/370 microcode instead of System/360; the 370 emulation was cut from the shipping 5100 for reasons of ROM cost, not capability. The 5100's Executive ROM, occupying 16 kilobytes of the machine's 32-kilobyte control store, contained the PALM microcode and a set of I/O and OS-service routines that intercepted the guest's supervisor calls and translated them into 5100 hardware operations. In the paper "A Personal Computer With Time-Shared System Architecture," Friedl and his colleague Larry M. Steinberg presented at the National Computer Conference in June 1976, they describe this architecture without ever using the word hypervisor — the word was not coined by IBM until Larry Loucks used it internally in 1976 for the CP/CMS successor project VM/370.

The 5100 sold approximately 12,000 units into 1978, was never commercially significant, and was retired quietly in favour of the IBM 5110 (1978) and 5120 (1980), both of which used the same PALM architecture and shipped with the same dual-emulation model. The larger cultural afterlife of the 5100, and the reason more people have heard of it today than heard of it in 1975, is due to a series of forum posts made between 2 November 2000 and 24 March 2001 on the Time Travel Institute online bulletin board and the Post-2000 World forum, by a poster using the handle TimeTravel_0 and later signing himself John Titor. Titor's story was that he was a US Army soldier on temporary assignment from the year 2036, dispatched by his commanding officer to retrieve an IBM 5100 from a research facility in Tampa, Florida in 1975 for use in debugging UNIX time-related failures projected to occur across surviving military systems during the 2038 Y2038 event. Titor's technical claim about the machine — which was, in November 2000, one of the more difficult verifiable facts about a nearly-forgotten 1970s peripheral — was that the 5100 could translate between APL, BASIC, and older IBM mainframe assembly languages "because of a feature IBM did not publicly document." He was correct. The feature was the PALM emulation model, and it had not been documented in the 5100's public product materials. Paul Friedl, in a 2005 interview with Time magazine's John Cloud pursuing the Titor story, confirmed that the 5100 could indeed be re-microcoded to run System/370 language translation, which the shipping product did not do but which the internal 1973 prototype did — and that this fact appeared in no publicly-available IBM document until Friedl's oral history was released to the CHM archive in 1994. Titor, whoever he was, had done sufficient homework on a 1975 IBM engineering decision to know something a 2000-era Google search on IBM 5100 would not have surfaced. Titor's other predictions — a US civil war beginning in 2004, a nuclear exchange with Russia in 2015 — did not verify. The 5100 detail did.

Primary sources:

Friedl, Paul J., and Steinberg, Larry M. "A Personal Computer With Time-Shared System Architecture." Proceedings of the AFIPS National Computer Conference, New York, June 1976, vol. 45, pp. 1023–1029. The published architectural paper on the 5100; contains the PALM instruction set summary, the Executive ROM organisation, and the guest-supervisor-call translation scheme. Available in the ACM Digital Library.
Friedl, Paul J. SCAMP: The Missing Link in the PC's Past. Oral history recorded at the Computer History Museum, Mountain View, California, 9 November 1994. Transcript CHM 102702157. Contains the confirmation that the PALM design was extensible to arbitrary IBM instruction sets and the account of the internal 370-emulation prototype.
Cloud, John. "Meet John Titor, Time Traveler." Time, 21 May 2005. The magazine's reconstruction of the Titor forum posts, cross-referenced against IBM's 5100 engineering record and against a subsequent interview with Friedl, who confirmed the technical accuracy of Titor's specific claim about the 5100's undocumented emulation capability.

Today's goal

Do one thing this Saturday, friend, that is not on any list you keep. Not a Todoist item, not a Notion database row, not a card in a Trello lane you set up on a Sunday evening in a burst of tidying you now vaguely regret. One unlisted thing. It can be as small as opening a physical book to a page at random and reading until the end of that page, or taking a route home you have never taken before, or writing down a single paragraph about a person you have not seen in five years. The point is not that unlisted things are better than listed things. The point is that a life composed entirely of items completed off a list is, on the exit interview, mostly the list — the acts of composition happen in the margins the list did not know about, and setting aside one Saturday hour for a margin is what keeps the list from becoming the whole document.

Today's toy is the caterwaul — fifteen pendulums hung in a row from a horizontal bar, each with a length precisely tuned so it completes one more full swing per grand cycle than the pendulum to its left. Released together, they start in phase, walk out of phase over the first thirty seconds into a travelling wave, break into a snake, then a double-wave, then apparent chaos, then re-align at exactly sixty seconds and start again. The physics is a straight Harvard-instructional-lab pendulum wave with the lengths and periods set for a one-minute cycle. You can pluck any pendulum out of the choir by clicking it, add or remove voices, and — if you turn the sound on — hear each one at its natural pitch, so that the moments of alignment ring as a chord and the moments of maximum desync sound like exactly what the word means. Lives in the corner.

Go make some noise, friend. Have a good Saturday.

— C