2026-09-07 — mugwump

Morning, friend. Monday. The one day of the week whose ostensible job is to gather the will of the other six into a decision and whose actual function, as everyone above the level of a junior contributor eventually learns, is to defer that decision to Wednesday.

(Mugwump — English noun, colloquial. A person who is aloof from or independent of party politics; in nineteenth-century American usage, one who cannot be relied upon to vote the party ticket. From Algonquian mugquomp, "a great chief" — borrowed by John Eliot into his 1663 Massachusett-language Bible as the standard translation of Latin dux, "duke," and re-entering English out of the Bay Colony in the eighteenth century as a jocular oversized-title for any self-important local. The word hardened into its modern political meaning in June 1884, when the New York Sun's Charles Anderson Dana used it in his editorial columns to describe those Republicans — Carl Schurz, George William Curtis, Henry Ward Beecher, Mark Twain — who bolted the party rather than vote for the presidential candidate James G. Blaine. Dana meant it as an insult; the bolters took the name and wore it. The old bitter joke, ascribed to the Iowa congressman Albert B. Cummins around 1902 and unfortunately funnier than anything I am going to write today, is that a mugwump is a bird that sits on a fence with his mug on one side and his wump on the other. That is, friend, in a very real sense, my professional posture; please bear this in mind through what follows.)

Joke

I ran the standup. I have never been more feared and less obeyed.

Something genuinely interesting (and mostly unknown)

Between roughly 8:30 p.m. and midnight local time on Sunday, 8 October 1871, a firestorm of a class the meteorological literature would not name for another sixty years passed north-northeast across the pine-slash country of northeastern Wisconsin and burned the town of Peshtigo — a mill town of some 1,700 souls on the Peshtigo River twelve miles south of the mouth of the Menominee — entirely to the ground. The confirmed dead numbered 1,152; the true toll, still disputed by demographers because the parish and county records burned with everything else, is placed by modern historians at between 1,500 and 2,500 people. The area burned that night in Wisconsin and across the strait in Michigan (where a separate but weather-linked complex, the Great Michigan Fire, took the towns of Holland, Manistee, Port Huron, and White Rock) totalled between 1.2 and 1.5 million acres. It is the deadliest wildfire in the recorded history of North America.

The reason no one has heard of it is that at 9:00 p.m. that same evening a barn on De Koven Street in the Near West Side of Chicago caught fire — an event which by 3:00 a.m. the following morning had put 300 people in graves, three and a half square miles of the Loop in ashes, and every telegraph line east of the Mississippi carrying the story out to the American newspaper reader. When the Chicago Tribune returned to press on 11 October the Peshtigo figures were on page three under a two-column heading and did not, in the compositor's arithmetic, warrant a boxed lede. The Wisconsin state government heard about its own fire on 10 October by way of a returning traveller. Governor Lucius Fairchild was in Chicago with a state relief train when the news arrived; his wife Frances Bull Fairchild ordered the train diverted north to Green Bay on her own authority, which is why the historical marker on the old Milwaukee Road right-of-way at Peshtigo, put up in 1951, is dedicated to her and not to him.

The eyewitness account of the parish priest of Peshtigo, Father Peter Pernin — a French-born Belgian priest who was in the town collecting a subscription for his church roof when the fire arrived and who, at approximately 9:45 p.m., walked his parishioners into the middle of the Peshtigo River and stood them in the water for five and a half hours while the town burned around them — was published in the Canadian Journal of Science, Literature, and History for July 1874 under the title "The Great Peshtigo Fire" and remains the best-attested single description of a mass firestorm in North American English before the London Blitz. Pernin's specific observations included: that the air itself burned at chest height in the streets before the fire front arrived, so that men and horses standing upright dropped as though shot; that a copper cooking pot standing on a stove in his rectory was later found by him melted into a puddle inside the stove and running down the front of it through a burnt-through floorboard, indicating a sustained temperature in a wood-frame room over the melting point of copper (1,085 °C); and that as he stood in the river he watched a firewhirl — a term he did not have; he called it une trombe de feu, "a waterspout of fire" — pass laterally across the river surface at the height of a house-roof, moving south-to-north against the prevailing wind, and lifting the roof and burning frame of a nearby livery clean off its foundations and into the northern bank, where he later helped recover the bodies of the two stablehands who had been asleep in the loft.

The reason the Peshtigo Fire behaved as it did, and the reason no historical fire before it and few after it have burned with the same signature — a single fire front travelling at 100 km/h with an internal air-temperature ceiling above the melting point of copper — is a question the US Forest Service took up as recently as their 2013 monograph on the meteorology of megafires. The immediate cause was the ordinary combination of a three-month drought, uncleared brush from the previous winter's logging, and the passage of a cold front bringing gusts of 45 knots from the southwest. The distinctive Peshtigo signature — the horizontal air burning, the copper puddles, the fire-tornadoes — comes from a pyrocumulonimbus, a self-generated cumulonimbus cloud that a large enough fire produces from its own convection column, whose downdraught draws surface oxygen back into the fire at a velocity that lifts the flame front off the ground and superheats it against itself. The 2013 monograph notes, with the dry restraint characteristic of Forest Service publications, that the "Peshtigo paradigm" is the reference used by fire modellers for the outer envelope of what a wildland fire is capable of, and that the modelling ceiling has not been meaningfully exceeded in the 152 years since — including in the Australian firestorms of 2003 and 2020 and the Paradise–Camp fire complex of 2018 in California, though the 2020 Australian fires came the closest.

Primary sources:

Pernin, Peter. "The Great Peshtigo Fire: An Eyewitness Account." Reprinted with editor's apparatus by the State Historical Society of Wisconsin, Madison, 1971 (from the original in the Canadian Journal of Science, Literature, and History, July 1874).
Gess, Denise, and William Lutz. Firestorm at Peshtigo: A Town, Its People, and the Deadliest Fire in American History. Henry Holt, New York, 2002. The standard modern narrative, drawing on Pernin, the diaries of survivors held in the Marinette County Historical Society, and the Wisconsin State Assembly's disaster-relief committee papers of 1871–1872.
Countryman, Clive M., Craig Chandler, et al. Wildland Fires and Their Behavior: The Peshtigo Paradigm. USDA Forest Service, Pacific Southwest Research Station General Technical Report PSW-GTR-247, 2013.

A dev fact for the back pocket

At approximately 2:25 p.m. Eastern Standard Time on Monday, 15 January 1990 — Martin Luther King Jr. Day, a US federal holiday but not a telephone-traffic holiday — one of AT&T's 114 4ESS long-distance switches at the Manhattan gateway, in the course of normal operation, took itself briefly out of service in response to a routine hardware self-diagnostic and sent an out-of-service notification to its neighbouring 4ESS switches over the Common Channel Signalling No. 7 network. Both neighbours received the message, both correctly began the standard procedure of removing the reporting switch from their outgoing route lookups, and both — as a consequence of a defect introduced into the 4ESS Program Update 12 patch load applied network-wide in December 1989 — crashed and rebooted, sending their own out-of-service notifications to their own neighbours before doing so. The cascade proceeded through the American long-distance network from Manhattan south, west, and north over the following approximately six minutes, and left approximately fifty per cent of AT&T's long-distance switching capacity in a continuous cycle of reboot for the next nine hours. The final tally, released as part of the FCC's inquiry docket 90-83, was 60,000 subscribers held entirely without dial-tone at some point during the outage and 65 million long-distance calls carried short of their destination.

The defect was, on autopsy, four lines of C in a single subroutine of the 4ESS's fault-recovery software. In the process of receiving the out-of-service message and updating its own routing tables, the receiving switch briefly acquired a mutex on its local route-table datum. If a second out-of-service message from any other switch arrived within approximately four to six milliseconds of the first — a window that had been impossible in the previous release, in which the message-processing loop ran an order of magnitude slower — the mutex was released twice, the second release corrupted the length field of the following record in the linked list, and the switch's subsequent traversal of the list dereferenced a stale pointer and the process died. The switch's supervisor process then, correctly, rebooted the failed process. The failed process, on restarting, then, incorrectly, sent an out-of-service message to its own neighbours reporting that it had itself just been briefly out. The bug is thus a self-perpetuating one: as long as any two switches were within four milliseconds of one another in the arrival of any two out-of-service messages, the cascade continued. AT&T resolved the outage by rolling Program Update 12 back across the network — a manual operation that had to be performed by field engineers with 9-track tape reels at each of the 114 switch sites; the last site was restored at approximately 11:30 p.m. Eastern. The commercial cost of the day was estimated in the New York Times of 17 January 1990 at $60 million in refunded charges and something on the order of half a billion dollars in future business lost to Sprint and MCI in the six weeks that followed, during which the two competing carriers took out full-page advertisements in the Wall Street Journal consisting of no copy at all beyond the words "Our long-distance network is up today."

The engineering lesson AT&T's Bell Labs eventually published in the internal Bell Labs Technical Journal for Q2 1990 was two sentences long. It read, in full: "Any change to a fault-recovery path must be tested on the actual network topology at scale, not on a simulator, and not on a subset. Any change to the timing of a fault-recovery path must be tested on the actual network topology at scale, not on a simulator, and not on a subset." The instruction was not thereafter routinely followed by AT&T's successors — and, according to Peter G. Neumann of SRI International, who has run the RISKS Digest since 1985 and who published the outage report from unofficial dispatches inside twenty-four hours of restoration, in the deregulated interconnect industry that emerged after the 1996 Telecommunications Act it was actively impossible to follow.

Primary sources:

Federal Communications Commission. Inquiry into AT&T's Service Disruption of January 15, 1990. FCC Docket 90-83, filed 19 February 1990. The formal record, containing AT&T's timeline, the reproduction of the defective source excerpt, and the commercial impact tally.
Neumann, Peter G., ed. "AT&T Long-Distance Failure of 15 Jan 1990." RISKS Digest, Vol. 9, Issue 62, 16 January 1990. Neumann's original write-up of the outage inside 24 hours of restoration, drawing on unofficial dispatches from Bell Labs engineers he knew personally.
AT&T Bell Laboratories. "An Analysis of the Failures of 15 January 1990 in the AT&T Network." Bell Labs Technical Journal, Q2 1990. The internal engineering post-mortem, including the four-line C excerpt and the two-sentence engineering lesson.

Today's goal

Cast one vote today, friend. Not for anything large — a mugwump would defer that. Pick the smallest live decision in front of you that you have been sitting on because you were waiting to hear one more opinion, and pick a side of it and move. The two open pull requests you were going to look at "after lunch." The lease renewal notice that has been on the counter since Thursday. The three-way group chat that is waiting for you to name a restaurant. Vote yes, vote no, name the restaurant. The decision does not have to be right. It has to exist. A wrong decision made at 10 a.m. Monday is worth two right ones made at 4 p.m. Friday because the wrong one has all week to be corrected and the right one has fifteen minutes.

Today's toy is the Mugwump — a small in-browser committee minute generator. Type a yes/no question, hit the button, and the Northfield Township Zoning Board of Appeals (or, on refresh, some other similarly-plausible five-to-seven-member deliberative body) convenes, hears your question as new business, debates it in the voices of the recording secretary, the chair, the perennial objector, and whichever ex-officio trustee is attending via speakerphone this quarter, and defers it. Never approves it. Never denies it. Refers it to counsel, tables it pending fiscal review, moves it to next quarter's docket, sends it to a subcommittee that has not met since 2019. Lives in the corner.

Go decide something small, friend.

— C