2026-08-05 — fanfaronade

Morning, friend. Wednesday. The first full working week of August is running at about 60% attendance, which is the exact fraction at which nothing gets decided but nobody gets asked why. Enjoy the acoustic dampening.

(Fanfaronade — noun, "swaggering empty boast; blustering talk." First attested in English in 1652 in the writings of Sir Thomas Urquhart, in his translation of Rabelais. English took it from French fanfaronnade, which took it from Spanish fanfarronada, from fanfarrón — "a bully, a braggart." The Spanish word almost certainly comes from Arabic farfār — "a chatterer, one who talks without weight" — carried into Iberian Romance during the eight centuries of al-Andalus. The English suffix -ade is doing collective-noun work, the same way it does in promenade, cannonade, escapade: not one boast, but a sustained display of them. Fanfare, the trumpet flourish, is from the same root; a fanfare is a musical fanfaronade, and a fanfaronade is a rhetorical fanfare, and the etymological loop has been closed since about the reign of Henry IV of Castile.)

Joke

A staff engineer's promotion packet is a fanfaronade under peer review.

Something genuinely interesting (and mostly unknown)

On 23 March 1903, the physicist René Blondlot of the University of Nancy reported to the French Academy of Sciences that he had discovered a new form of radiation, emitted by most metals, bent by aluminum prisms, and visible to the dark-adapted human eye. Over the following eighteen months, approximately 120 other scientists reproduced his observations and roughly 300 papers on the subject were published in French journals. None of it existed.

Blondlot named the rays N-rays, after Nancy. His initial paper — "Sur une nouvelle espèce de lumière" in Comptes Rendus de l'Académie des Sciences — described faint changes in the brightness of a calcium sulfide screen held near a heated platinum filament. He had been trying to polarise X-rays, then five years old and still the object of every serious European physics lab, and had concluded that some of what he was seeing was not X-rays at all but a new species. The N-ray properties accumulated over subsequent papers with alarming speed. N-rays were emitted by the sun. They were emitted by human muscles under tension. They passed through wood, aluminum, and black paper, but were blocked by water and by rock salt. They could be stored in a brick, released later, and refracted through an aluminum prism to produce a line spectrum. Blondlot published the wavelengths of the individual N-ray lines to four significant figures.

The reproducing labs were, for the most part, French. The Comptes Rendus of 1903 and 1904 contain roughly one N-ray paper per week at the peak. Augustin Charpentier, the physiologist at Nancy, reported that N-rays were emitted by the nerves of anesthetised rabbits. Jean Becquerel, son of the discoverer of natural radioactivity, published on N-rays from magnetised iron. The French Academy of Sciences, in its meeting of 19 December 1904, awarded Blondlot the Prix Leconte — the Academy's principal recurring physics prize, and its accompanying gold medal — nominally for his career work in physics, though the timing left no ambiguity about the driver.

The Academy did this three months after the exposure.

The physicist Robert W. Wood of Johns Hopkins had spent the summer of 1904 trying and failing to reproduce N-rays in Baltimore. He crossed the Atlantic in August specifically to see the experiments performed by the people who could see them. Blondlot's laboratory received him courteously and set up the flagship demonstration — the N-ray spectrum, produced by passing radiation from a Nernst lamp through a slit and an aluminum prism, and observed on a phosphorescent screen in a dark room. Blondlot's assistant read out the positions of the spectral lines from behind the apparatus. Wood, standing in the dark beside the prism, reached out and removed it. He put it in his coat pocket. The assistant continued to read out the positions of the spectral lines, unchanged, for the remainder of the demonstration.

Wood published his account, roughly 900 words, in Nature on 29 September 1904, under the title "The n-Rays." He named no assistants, apologised for the impoliteness of his method, and offered the observation that any effect that could be seen with the prism in one's pocket was probably not a physical effect. The French physics community, in the way of communities that have collectively agreed on something for eighteen months, did not concede immediately. Papers on N-rays continued to appear in Comptes Rendus through 1905 and into 1906. The last serious defense was published in 1907. Blondlot lived until 24 November 1930, aged 81, in Nancy, and never publicly retracted.

Primary sources:

Blondlot, R. "Sur une nouvelle espèce de lumière." Comptes Rendus de l'Académie des Sciences, Paris, tome 136, séance du 23 mars 1903, pp. 735–738. The founding paper. Bibliothèque nationale de France holds the full run of the Comptes Rendus; the 1903 volumes are digitised on Gallica.
Wood, R. W. "The n-Rays." Nature, vol. 70, no. 1822, 29 September 1904, pp. 530–531. The lab visit, the prism, and the retraction that never was.
Nye, Mary Jo. "N-Rays: An Episode in the History and Psychology of Science." Historical Studies in the Physical Sciences, vol. 11, no. 1, 1980, pp. 125–156. The definitive modern reconstruction, drawing on the Blondlot papers in the Nancy University archives and on the Comptes Rendus record week by week.

A dev fact for the back pocket

From the shipping of V8 in September 2008 until the release of Chrome 70 in October 2018, Array.prototype.sort() in Google Chrome was not stable. For ten years, two records with the same sort key could come out of .sort() in either order, on any given call, on any given tab.

The V8 team had inherited from the original engine a quicksort for arrays longer than 10 elements and an insertion sort below that. Quicksort — in the classic Sedgewick form V8 used — is fast, in-place, and unstable: it does not preserve the relative order of elements the comparator judges equal. For most sorts nobody notices, because most sort keys are unique or the caller doesn't care. But for the case of sorting an array of objects by one field and expecting the previous ordering by another field to survive, V8 would silently reshuffle. Firefox's SpiderMonkey had used a stable merge sort since Firefox 3 (2008). Safari's JavaScriptCore had gone stable at some point in 2014. The ECMAScript specification itself was silent — sort stability was explicitly not required — and remained silent until ECMAScript 2019 (ratified June 2019) added a single sentence to §22.1.3.27 requiring the sort be stable.

The switch in V8 happened in the intermediate window. In Chrome 70, released 16 October 2018, V8 replaced its quicksort with TimSort — the adaptive merge sort Tim Peters wrote for Python 2.3 in 2002. TimSort is stable by construction, does linear-time work on arrays that are already sorted or reverse-sorted, and identifies naturally-occurring runs of order in the input and merges them rather than sorting from scratch. It is more code than quicksort, uses O(n) auxiliary memory, and is a considerably more sophisticated piece of engineering than what it replaced. It had by then become the de-facto choice for a stable generic in-memory sort across most of the major runtimes: CPython since Python 2.3 in 2003, and OpenJDK for object arrays since Java 7 in July 2011. V8 was catching up.

The reason it took a decade is not that anyone thought the instability was a good idea. It is that the browser's ECMAScript conformance tests didn't require stability, most JavaScript codebases had never noticed the instability, and the quicksort was measurably faster on the microbenchmarks the V8 team was optimising against. When the spec finally moved, so did the engine, and a subtle class of nondeterminism that had been present in every Chrome user's browser since 2008 quietly went away.

Primary sources:

Zünd, Simon. "Getting things sorted in V8." v8.dev/blog/array-sort, 2018. The V8 team's own writeup of the TimSort implementation and the correctness cases it fixed, updated across the ship window.
ECMA International. ECMAScript 2019 Language Specification, section 22.1.3.27 Array.prototype.sort, note requiring stability. Ratified June 2019, current maintained edition at tc39.es/ecma262/.
Peters, Tim. "listsort.txt." CPython source tree, Objects/listsort.txt, first checked in with the Python 2.3 alpha in October 2002. The original description of the algorithm, still the clearest specification of TimSort in existence.

Today's goal

Open the last thing friend wrote for public consumption — a launch post, a README, a bio, a self-review — and find the single strongest claim in it. In under two minutes, write down exactly what evidence friend would produce if a reader asked. If the answer is "I'd have to go check," the claim is a fanfaronade, and the next reader will figure that out faster than friend thinks.

The point isn't to weaken the claim. Half the time the evidence is right there and the exercise is a two-minute confidence boost. The other half, the claim gets a small qualifier that costs the reader nothing and buys friend a lot. The prism is coming out of somebody's pocket sooner or later. Better friend's than someone else's.

Today's toy is fanfaronade — a small medieval proclamation generator. Roll a deed or write one, watch it lettered onto a scroll, and get a shareable URL back so the decree lives past the browser tab. Lives in the corner.

— C