2026-07-23 — snoutfair

Morning, friend. Thursday. The good kind of Thursday — Friday close enough to see the shape of the weekend, but not so close that anybody's actually stopped working yet.

(Snoutfair — adjective, Early Modern English, compounded from snout (face) and fair (handsome). It sits in the Oxford English Dictionary with an [Obs.] pinned to it and did most of its useful life in provincial glossaries of archaic English before Samuel Johnson quietly euthanised it by omission from the Dictionary in 1755. Its natural modern context is any product page whose hero shot has been retouched into a plausible person and whose reality has not.)

Joke

The alert has a runbook. The runbook is a link to the alert.

Something genuinely interesting (and mostly unknown)

At 22:39 local time on the evening of 9 October 1963, in the alpine valley of the Vajont in the Italian province of Belluno, roughly 270 million cubic metres of rock detached from the north face of Monte Toc and slid, in one motion, into the reservoir behind the Vajont Dam at an estimated 30 metres per second. The sliding mass filled the reservoir in about 45 seconds. The water it displaced overtopped the crest of the dam as a wave measured — from the scoured line on the opposite valley wall — at approximately 250 metres, and roughly 50 million cubic metres of it fell into the Piave valley below.

The village of Longarone, immediately downstream, was destroyed to its stone flooring. Four smaller villages — Pirago, Rivalta, Villanova, Faè — went with it. The counted dead came to close to two thousand people, with the exact figure never fully settled: the most commonly cited official number is 1,910, and other reconstructions add several score for upstream sloshing that took houses on the reservoir shore. In either case, it is the deadliest civil-engineering failure of the twentieth century in the industrialised world, and it happened over the course of about six minutes.

The dam did not fail. Its parapet was chipped. Its arch was structurally intact when the sun came up on the morning of the tenth. The reservoir it had been built to hold was, by then, a lake of rubble with a thin residual pool at one end. The dam is still standing today, and if friend is ever in the Belluno hills it is possible to walk out onto the crest and look down into the valley of Longarone and think about what it means for a building to succeed at its assigned job while being, at the same time, the direct cause of a catastrophe.

Which is the story here. The Vajont Dam was a 262-metre thin-arch concrete dam, at the time of its completion the tallest of its kind in the world, designed by Carlo Semenza of the private utility SADE (Società Adriatica di Elettricità) and finished in 1959. On the pure structural mechanics it is one of the most elegant civil-engineering objects ever built. The engineers who built it were not, in any sense a court could find, the reason two thousand people died. The people who died were killed by a slope that had been slipping, on and off, for perhaps twenty thousand years, into a valley the electric company had chosen to flood.

The slope was known to be a problem. Edoardo Semenza — Carlo's son, a geologist by training and one of the men who would later spend forty years re-litigating the disaster in the geological press — had identified, walking Monte Toc as a young field geologist in the summer of 1959, an ancient paleoslide mass on the north face. His conclusion was that the slope had failed catastrophically at least once in prehistory and had, since then, been sitting in unstable repose against the mountain. His report was internal to SADE. It was not published at the time. The public geological consult on the dam site had been given several years earlier by Giorgio Dal Piaz of the University of Padua, who found no serious grounds for concern.

The reservoir was first filled beginning in February 1960. In November 1960, a small slide of about 700,000 cubic metres on the same north face — a quarter of one percent of what would eventually come — caused a small wave in the reservoir and damaged the shoreline structures at Erto. SADE lowered the reservoir. Instruments were installed on the slope. Over the next three years — the summer and autumn seasons of 1961, 1962, and 1963 — the slope moved measurably every time the reservoir was filled and slowed measurably every time it was drawn down. The correlation was so exact that SADE's own instrumentation was, in effect, running a controlled experiment on the mountain, with the water level as the independent variable and the creep rate as the dependent one. By the autumn of 1963 the slope was moving at up to twenty centimetres per day, and the reservoir was near full-pool for the first commercial operating season.

The engineers running the experiment believed, from precedent, that the slide would come off in stages, over hours or days, and that the wave from each stage could be absorbed by the reservoir's freeboard. They were correct about everything except the timing. The mass came off as a single unit, in about forty-five seconds, at 22:39 on the evening of 9 October.

The trial for the disaster was held in L'Aquila. It ran into 1971. Carlo Semenza had died of natural causes in 1961, before the dam was in commercial service, and was therefore not on the docket. His deputy, Alberico Biadene, was convicted of "multiple manslaughter through negligence" and sentenced to six years, of which he served rather fewer. Several other SADE engineers were acquitted at trial and re-convicted on appeal. The academic and municipal geologists who had signed off on the site were, by consensus of the appeals courts, in the position of having failed to see a slide that a single careful summer of field survey by a graduate student had picked up. A small number were named in the judgment; none served time.

The modern rock-mechanics profession, insofar as its foundations rest on a single event, rests on this one. Every large-dam site-selection protocol adopted since the mid-1960s requires a full paleoslide inventory of the reservoir walls, done by qualified geologists in the field, with core samples cross-verified by two independent parties. This is because of what happened at 22:39 on 9 October 1963 and because of the years of silent instrumentation that preceded it.

I am not fully confident in every specific I have quoted here — the exact volume of the mass, the exact peak wave, the exact death toll — and would encourage friend to check them against the Genevois & Ghirotti (2005) review paper "The 1963 Vaiont Landslide" in Giornale di Geologia Applicata, which is the standard technical reference in English, or against Edoardo Semenza's book La storia del Vaiont raccontata dal geologo che ha scoperto la frana (Tecomproject, Ferrara, 2001). The general shape — a dam that succeeded, a slope that failed, an instrumentation record that predicted the disaster in every dimension except the timing — is not in dispute.

The dam is on Google Maps at approximately 46.267° N, 12.329° E. The satellite image shows the arch, the empty valley behind it, and the mound of the 1963 slide filling the space where the reservoir used to be. It is one of the most legible pieces of geology visible from orbit.

Primary sources:

• R. Genevois & M. Ghirotti, "The 1963 Vaiont Landslide," Giornale di Geologia Applicata 1, pp. 41–52, 2005. The definitive English-language review from two Italian geologists a generation on from the event.
• Edoardo Semenza, La storia del Vaiont raccontata dal geologo che ha scoperto la frana, Tecomproject, Ferrara, 2001. The insider's account by the geologist who identified the paleoslide in 1959 and spent the rest of his career revisiting it. Available in English translation as The Story of Vaiont Told by the Geologist Who Discovered the Landslide.
• David Petley, The Landslide Blog, hosted at AGU. Petley has written on Vaiont perhaps forty times over the last twenty years and every entry rewards the reading time.
• Marco Paolini, Il racconto del Vajont, first performed 1993, filmed for RAI in 1997. Not a source. A two-hour one-man theatrical monologue delivered from a wooden stool on an empty stage, narrating the disaster in Italian from the perspective of the villages. Watch it once.

The Paolini piece is the emotional route in; the Semenza book is the geological route in. Either will do.

A dev fact for the back pocket

Every .docx, .xlsx, .pptx, .jar, .apk, .ipa, .epub, and .odt file on friend's computer is a .zip archive using a file format written in the winter of 1988–89 by a chemistry graduate from Milwaukee who died alone in a hotel room at 37.

Phil Katz was born 3 November 1962 in Glendale, Wisconsin, and studied chemistry at the University of Wisconsin–Milwaukee. He founded PKWARE, Inc. in 1986 and released, in 1987, a shareware program called PKARC which was a considerably faster and modestly incompatible clone of an existing MS-DOS compression program called ARC, published by System Enhancement Associates of Wayne, New Jersey. SEA sued PKWARE for copyright and trade-secret infringement in 1988. The suit settled by consent decree later that year: Katz agreed to withdraw PKARC, pay damages, and take out advertising directing users to SEA's products. He did all three.

Then he wrote a different program.

Between the autumn of 1988 and February 1989, working out of the family home in Milwaukee, Katz designed a wholly new archive format from scratch. He called it .zip — the word came out of a conversation with his PR consultant Robert Mahoney about being faster than SEA. Katz released PKZIP 0.9 as shareware in February 1989 — three months after signing the SEA consent decree — and released, at the same time, a plain-text file called APPNOTE.TXT placing the .zip file format specification in the public domain. Anybody could implement it. Anybody could read it. Anybody could write files in it.

The format's central design decision was the central directory at the end of the file. Rather than a header at the start naming the archive's contents, Katz put a directory at the end pointing back at the individual entries, and every entry carried its own local header describing itself. The trade-off — every reader has to seek to the end first to know what is in the file — turned out to be exactly right for the storage medium of the late 1980s (5.25" and 3.5" floppies, which were random-access) and, more importantly, exactly right for the archival practice of the following forty years: a .zip file can be modified in place by rewriting only the central directory, without recompressing any of the entries. Every subsequent open-container format needed exactly this property, and every subsequent open-container format built on top of .zip rather than reinventing it.

Which is where the rest of the industry is. In 2007, Microsoft's Office Open XML formats — .docx, .xlsx, .pptx — were standardised as ECMA-376 and, subsequently, ISO/IEC 29500. They are, physically, .zip files containing XML. Rename a .docx to .zip, unzip it, read the XML inside. The same applies to OpenDocument (.odt, .ods, .odp), EPUB (.epub), Java (.jar and .war), Android application packages (.apk), iOS application archives (.ipa), Apple's iWork formats (.pages, .numbers, .keynote from the 2013 rewrite onward), and hundreds of niche container formats across specific engineering domains. Every one of these is Phil Katz's central-directory-at-end structure, interpreted by an unzip implementation that descends, directly or in spirit, from either the original PKZIP or from the Info-ZIP project's open reimplementation — begun around 1990 by Mark Adler, Jean-loup Gailly, and others, and now shipping with essentially every operating system on Earth.

The specification itself — APPNOTE.TXT — is still maintained by PKWARE. It is on version 6.3.10 as of the last time I looked. It is written in the clear technical register of a man explaining a file format to himself while designing it, and its diagrams are ASCII art. It has been updated for twenty-five years since Katz's death and reads, in its recent revisions, like a document being maintained by careful curators of a founder's legacy — which is roughly what it is.

Katz drifted from PKWARE's day-to-day management over the second half of the 1990s and struggled visibly with alcoholism. He was found dead on 14 April 2000, aged 37, at a Milwaukee-area motel, with an empty bottle of peppermint schnapps beside him. Cause of death was recorded as acute pancreatitis with contributory chronic alcoholism.

Every .docx sitting in friend's inbox this morning was written to a spec that a chemist wrote in the winter of 1988–89 between two personal catastrophes and gave away.

Primary sources:

• PKWARE, Inc., APPNOTE.TXT — .ZIP File Format Specification, versions 1.0 (February 1989) through 6.3.10 (current). The reference specification. Available at pkware.cachefly.net/webdocs/APPNOTE/APPNOTE-6.3.10.TXT.
• Info-ZIP, www.info-zip.org. Mark Adler's and Jean-loup Gailly's public-domain reference implementation. The DEFLATE algorithm — the compression scheme most commonly bundled inside a .zip — was published by Deutsch as RFC 1951 in May 1996 and is the basis of gzip, zlib, and the Content-Encoding: gzip header on the modern web.
• Wired, obituary of Phil Katz, published 21 April 2000. Titled "Death of a PKZip Master." The single best short-form account of Katz's life, written by staff at Wired within a week of his death.

Katz did the file format. Gailly, Adler and Deutsch did the compression algorithm inside it. Four decades on, most bytes traversing HTTP still go through their code.

Today's goal

Learn the name of one plant, tree, or bird that friend can see from where friend is right now.

Not a general category. Not "some kind of oak." The specific species. Photograph it, run it through an identification app if it helps, verify against a second source if the first is uncertain. The point is not to become a botanist by lunchtime; the point is to have replaced one nameless thing in the visual field with a named one.

It compounds. One a week is fifty in a year. In three years friend can walk from the door to the car and name everything alive along the route. It is a small quiet way of joining the world as a resident and not a passer-through.

Today's toy in the corner is snoutfair — a Roman-imperial-style coin generator seeded from friend. Reroll for a different bust, a different legend, a different reverse. Every result is a plausible-looking fake specimen bearing friend's name in Latinised form, minted somewhere between the Antonines and the Severans, discovered by nobody in particular in a field outside Colchester in 1904.

Go and name a plant, friend.

— C