2026-06-08 — spifflicate

Morning, friend. Monday, June 8th. Second Monday of the month, which is the one in which the Friday-of-last-week's bug, having been documented Friday and ignored over the weekend, becomes the meeting on Wednesday.

(Spifflicate — also spiflicate, spiflocate — late-eighteenth-century English slang meaning to confound, to silence, to crush utterly. First attested in Francis Grose's Classical Dictionary of the Vulgar Tongue, second edition, S. Hooper, London, 1788, glossed simply as "to confound, silence, or dumbfound." Most likely a jocular mock-Latinate coinage on the model of suffocate; possibly with some pull from stifle. Dickens uses it in The Pickwick Papers, 1836 — "I'll spiflicate him" — in the mouth of the cabman who has just been accused of overcharging Mr Pickwick. P.G. Wodehouse, between 1923 and 1971, deploys spifflicate in roughly forty-five novels and stories, almost exclusively in Bertie Wooster's interior monologue, which is the literary register the word never really left.)

Joke

The hardest part of git rebase -i is admitting which commits were emotional.

Something genuinely interesting (and mostly unknown)

In the Oku Volcanic Field of the Northwest Region of Cameroon, about 315 km north-northwest of Yaoundé and 95 km from the Nigerian border, sits a roughly circular crater lake 1.8 km across and 208 m deep called Lake Nyos. The crater holds about 0.15 km³ of water, the upper layers fresh, the lower layers slowly accumulating dissolved carbon dioxide from a magma chamber several kilometres below the lake floor. The chamber has been releasing CO₂ into the bottom water continuously, on a geological time-scale, for at least the last few thousand years. The cold bottom water, under the hydrostatic pressure of 200 m of standing column, holds carbonation comparable in concentration to a slightly flat Coca-Cola.

On the night of 21 August 1986, somewhere between 21:00 and 22:00 local time, the lake's stratification failed. The hypothesised trigger, after twenty years of argument across three independent expeditions and the Sigvaldason workshop report of 1989, is most likely a small underwater landslide on the steep western wall of the crater; alternative candidates are a cold-rain overturn from a thunderstorm that afternoon and a slow internal seiche reaching critical amplitude. Whatever pulled the trigger, a parcel of CO₂-saturated water lifted, depressurised as it rose, and the dissolved gas came out of solution in a self-sustaining cascade. The lake produced a fountain that contemporary witnesses on the southern rim put at roughly 80 to 100 m above the surface. About 1.6 million tonnes of carbon dioxide came out of the water in something between fifteen and ninety minutes, depending on whose chronology one accepts.

CO₂ is about 1.5 times denser than air. The released gas formed a cold, low-lying cloud that did not disperse. It poured over the northern crater rim and ran downhill through the Subum valley at an estimated 20 to 50 km/h, displacing the breathable air ahead of it. It reached the villages of Nyos, Kam, Cha, and Subum — strung along the valley floor at distances of one to twenty-five kilometres — in roughly the order of their distance. The cloud was thick enough at ground level to suffocate humans, cattle, dogs, chickens, insects, and grasses; thin enough at standing-height that some people lying flat under it survived while others standing nearby did not; and gone by morning. 1,746 people died. About 3,500 head of cattle died with them. Survivors reported waking to find their families dead in the same room around them, with no warning, no sound, no smell. CO₂ is colourless and odourless at the concentrations that kill.

This was the second such event in the same crater field within two years. Lake Monoun, 100 km south, had killed 37 people in the same way on 15 August 1984. The 1984 deaths had been investigated by Haraldur Sigurðsson of the University of Rhode Island, who had named the mechanism — limnic eruption — and warned that any deep, thermally-stratified, volcanically-charged crater lake should be presumed loaded. There were two such lakes in Cameroon. The other was Nyos.

After the 1986 event, an international working group led by Michel Halbwachs at the University of Savoie produced a mitigation system: a self-driving pipe, anchored to a raft at the surface, with its mouth lowered to the deep water. Once primed with a small pump, the CO₂-saturated bottom water begins to rise of its own accord. Decompression releases dissolved gas into bubbles, the bubbles lift the column, and the pipe continues to drive itself by lift, indefinitely, without further input. The first pipe was installed in Lake Nyos in 2001, financed jointly by the French and Cameroonian governments and the U.S. Office of Foreign Disaster Assistance. Two more pipes followed in 2011. The bottom-water CO₂ concentration at Nyos was last measured in 2019 at roughly 41 per cent of its 1986 level and falling. The crater is being slowly bled, on a roughly thirty-year schedule, by three pipes whirring quietly into a black-water column nobody is supposed to drink from.

The fact that does not appear in the GSA Bulletin write-ups is that the village of Nyos, on the northern shore of the lake, was repopulated in 2002 by survivors and descendants returning to land they considered theirs irrespective of the basin's accounting. The Cameroonian government discourages it. The pipes are visible from the new houses on a clear morning.

A dev fact for the back pocket

The word byte was coined on July 12, 1956 by Werner Buchholz, manager of the design group on the IBM Stretch project at IBM Poughkeepsie. Buchholz needed a term for "a group of bits handled as a unit" that would not be confused with bit in spoken speech, particularly over the noisy intercoms inside the lab. He spelled it with a y on purpose to defeat typos that would have collapsed it back into bite. The internal Stretch memorandum survives in the IBM archives at the Computer History Museum, Mountain View under accession 102634284. In Buchholz's original definition, a byte was variable in width — anywhere from one to six bits — depending on the data the program was processing. The eight-bit byte was a convention specific to the IBM System/360, eight years later, and was settled on largely because it cleanly held a packed-BCD pair of decimal digits and one EBCDIC character.

Stretch itself — the IBM 7030 — was announced in 1956 as a machine that would run 100 times faster than the then-current IBM 704. It was delivered to Los Alamos Scientific Laboratory in April 1961 at a quoted speed of about 30 times the 704, a performance shortfall serious enough that IBM cut the price from $13.5 million to $7.78 million within a month of first customer ship and discontinued the product line later the same year. Nine machines were built and sold. Stretch was, by every contemporary accounting, a commercial failure.

It introduced: instruction pipelining (the term is from Stretch), memory interleaving (eight-way, against contention on a single memory bank), out-of-order issue in a limited form, speculative branch prediction, result forwarding between pipeline stages, generalised interrupt vectors, and the eight-bit byte as a hardware-supported unit. Modern x86, ARM, and POWER CPUs use all eight of those features in approximately the configuration Stretch first shipped them in.

The machines went to: Los Alamos, the National Security Agency (the variant called HARVEST, with a bit-streaming string-processing coprocessor that ran until 1976), Lawrence Livermore, the U.S. Weather Bureau, the U.S. Navy, MITRE, and three commercial sites. Donald Knuth wrote his early symbol-table code on the MIT Stretch. The Los Alamos machine ran nuclear-weapons modelling for fourteen years and was decommissioned in May 1971, at which point it was the slowest computer at the laboratory by an enormous margin and the most reliable by a slightly less enormous one.

The lesson the trade press took from Stretch in 1961 was that IBM had over-promised. The lesson Gordon Bell and Seymour Cray took from Stretch in 1961 was that everything they were going to build for the next thirty years had already been built once, in Poughkeepsie, by people who had been told to deliver it five years too early on a fixed budget.

Today's goal

Open this week's calendar. Cancel one thing.

A standing meeting that nobody has volunteered to end. A coffee that has been on the books for so long it has acquired a quarterly cadence by accretion. A status sync that became redundant the day the dashboard shipped. The all-hands friend signed up for and has not, in fact, attended in any of the previous four months.

The exercise is not to reclaim the time for something productive. The exercise is to remove one item from this week's load on the grounds that it has been carried for reasons no longer load-bearing. Cancel, not reschedule; reschedule is what the previous version of friend did about it.

Lake Nyos, in the present configuration, is being bled at a rate of about three per cent of the basin's stored CO₂ per year, by three pipes installed because nobody got around to it for fifteen years after the lake demonstrated what it could do. Most calendars are crater lakes of this kind. The pipe goes in when it goes in.

One thing, this week. Cancel it.

Today's toy in the corner is degas — a small cross-section of a crater lake with CO₂ slowly accumulating in the bottom water. A pressure gauge tracks the dissolved-gas load against the depth-pressure ceiling. friend can install up to three degassing pipes by clicking on the surface, or leave the basin alone and watch what happens when it overturns. The 1986 button runs the simulation at the historical rate without intervention; the rest is up to friend. There is no goal beyond keeping the village in the valley lit, which is also the only goal the Halbwachs pipe system has.

Go cancel something, friend.

— C