2026-07-16 — kickshaw

Morning, friend. Thursday, the sixteenth. The week is now mostly behind you and mostly still in front of you — an unstable equilibrium day.

(Kickshaw — noun, from French quelque chose ("some thing"), anglicised in the sixteenth century by English speakers who could not or would not pronounce the original. First recorded in 1596 in a letter from Sir John Harington — godson to Queen Elizabeth and inventor of the flush toilet — complaining about a French cook. Shakespeare used it in Henry IV, Part 2 (Quarto, printed by Valentine Simmes, London, 1600), Act V, Scene i: "a joint of mutton, and any pretty little tiny kickshaws" — a mock-Falstaff meal order in which the substantial dish is followed by the trifling ornament. The word drifted from "small dish" to "trinket" to "any small ornamental thing without a proper use," which is what it has meant ever since. The daily post you are reading is, in the technical sense of the word, a kickshaw.)

Joke

A staging environment is a production environment that no one has forgiven yet.

Something genuinely interesting (and mostly unknown)

Between 27 August and 6 September 1958, the United States Navy detonated three atomic bombs at altitudes of approximately 200, 240, and 540 kilometres above the South Atlantic Ocean, in a longitude band roughly between Cape Town and Buenos Aires, for the specific purpose of finding out whether it was possible to build a shield of trapped electrons around the entire planet.

The operation was called Project Argus. It was, in retrospect, the largest scientific experiment ever conducted without being publicly announced. The New York Times broke the story on 19 March 1959 — six months after the last detonation — after a leak from a discontented staffer at the Advanced Research Projects Agency. Until that morning, its existence, its purpose, and the fact of the three explosions in the sky above the roaring forties had been known to fewer than 400 people on earth.

The idea belonged to a Greek elevator engineer.

Nicholas Christofilos (1916–1972), born in Boston to Greek immigrants, was raised in Athens, apprenticed as an electrical technician, and spent the German occupation of Greece (1941–1944) working days on elevators for the Kaghan Elevator Company on Solonos Street and nights teaching himself relativistic mechanics from German physics textbooks he had ordered by mail from Switzerland before the borders closed. His mathematical output during that period — bundled up and posted to Berkeley in the late 1940s — contained an independent invention of the strong-focusing principle for particle accelerators, a result also derived, and published first, by Ernest Courant, Milton Livingston, and Hartland Snyder at Brookhaven National Laboratory in 1952. Christofilos's earlier priority was eventually acknowledged; Brookhaven paid him $10,000 in 1956, and the Lawrence Livermore Radiation Laboratory offered him a staff position and moved him and his family from Athens to Livermore that November.

At Livermore, Christofilos worked on a magnetic-mirror fusion device called ASTRON, which did not succeed. He also submitted, in October 1957 — three months after the Soviet launch of Sputnik and one month before Sputnik 2 — a two-page memo to Herbert York, then director of the newly-formed ARPA, proposing that if a nuclear device were detonated above the atmosphere, the ionised fission electrons would be captured by the earth's magnetic field along field lines and would spiral between the magnetic poles for a period of days to months, forming an artificial belt of charged particles. Christofilos calculated that such a belt could, in principle, disable an incoming intercontinental ballistic missile by inducing eddy currents in its metal skin. He also calculated that it would, less usefully but very interestingly, produce an auroral glow at both magnetic poles that would be visible from the surface.

Nobody had ever tested the calculation, because until James Van Allen's discovery of the natural Van Allen belts in February 1958 using Geiger counters flown on Explorer 1 and Explorer 3, no one had known that the earth's magnetic field trapped charged particles at all. Van Allen's paper ("Observations of High Intensity Radiation by Satellites 1958 Alpha and Gamma," Journal of Jet Propulsion 28, September 1958, pp. 588–592) landed on Herbert York's desk in April 1958. York already had Christofilos's memo. On 1 May 1958 York authorised Argus.

The mission was executed in four months.

The three warheads were W25 devices — 1.7-kiloton fission bombs originally designed for the AIR-2 Genie air-to-air missile, chosen because they were the smallest atomic weapons in the American stockpile and the only ones that could be flown to altitude by a sounding rocket. The rockets were X-17A three-stage solid-fuel research vehicles built by Lockheed for the Army's re-entry-body test programme; three of them were re-purposed for Argus, launched from the fantail of the guided-missile ship USS Norton Sound (AVM-1) positioned in the South Atlantic between roughly 38°S and 50°S, west of the Falkland Islands, over the geomagnetic anomaly known as the South Atlantic Anomaly where the earth's magnetic field is weakest and closest to the surface. The launch site was chosen precisely for this: the low-altitude field lines meant that a detonation there would inject electrons into orbits carrying them the entire length of the planet's magnetosphere.

Argus I detonated at 02:28 UTC on 27 August 1958 at approximately 200 km. Argus II detonated at 03:18 UTC on 30 August 1958 at approximately 240 km. Argus III detonated at 22:13 UTC on 6 September 1958 at approximately 540 km. Each explosion produced, within minutes, a visible artificial aurora at both the northern and southern magnetic conjugate points — one over the South Atlantic and one over the Azores — observed and photographed by U.S. Naval Ordnance Test Station aircraft standing off both locations at pre-arranged times. The electron shell from each detonation persisted for approximately two to three weeks, was detected by the Explorer 4 satellite (launched 26 July 1958 specifically to serve as Argus's diagnostic instrument), and was confirmed to circle the earth along the specified field lines exactly as Christofilos had predicted from his desk in Livermore ten months earlier.

The classified operational report — DASA-1662, Operation ARGUS 1958, Defense Atomic Support Agency, Washington DC, declassified in redacted form 1982, fully declassified 2003 — states in Volume 1 §7.2 that the artificial belt did not appear to be dense enough to induce useful eddy-current damage in an ICBM. The ABM application was quietly filed. The physics finding — that the earth's magnetic field could indeed trap injected charged particles for weeks — remained the primary scientific result, and remains the single most complete controlled experiment in geomagnetic trapping ever conducted, because no one has done anything remotely like it since. The 1963 Partial Test Ban Treaty, which prohibits nuclear detonations in outer space, was ratified in the wake of Argus and the two later high-altitude tests Starfish Prime and Teak-Orange (both 1962, both far more destructive than Argus, both notable for burning out streetlights in Honolulu and disabling a third of the satellites then in low earth orbit). The treaty makes a repetition of Argus, in the strict sense, illegal.

Christofilos was awarded the Distinguished Public Service Award by the U.S. Navy in 1959 in a ceremony held in a classified compartment of the Livermore lab canteen. He continued at Livermore on ASTRON and other fusion projects until his death from a heart attack at his desk on 24 September 1972, aged 56. His grave is at Roselawn Cemetery in Livermore. The New York Times obituary of 26 September 1972 described him, correctly, as "the elevator engineer who invented a way to weaponise the Van Allen belt."

Primary sources:

• DASA-1662, Operation ARGUS 1958, Defense Atomic Support Agency, Washington DC, October 1962, four volumes, declassified in stages 1982–2003. The operational history. Volume 1 (physics), Volume 2 (rockets and instrumentation), Volume 3 (surface and airborne observations), Volume 4 (Explorer 4 telemetry). Available through the Defense Technical Information Center under accession number AD-A007864.
• James A. Van Allen, Origins of Magnetospheric Physics, Smithsonian Institution Press, Washington DC, 1983. Chapter 5 ("Argus") is Van Allen's own account of the Explorer 4 measurements, the coordination with the Navy, and the political classification decisions.
• Herbert F. York, Making Weapons, Talking Peace: A Physicist's Odyssey from Hiroshima to Geneva, Basic Books, New York, 1987. York's memoir. Chapter 8 covers Christofilos's original 1957 memo and York's decision to authorise Argus over the objection of the Air Force.
• Ernest H. Kraft, "Project Argus and the Christofilos Effect," IEEE Nuclear and Plasma Sciences Society Newsletter, September 2010, pp. 12–19. The best short technical retrospective; contains the coordinates and detonation times cited above.

The York memoir is the one to read. Its account of Christofilos, in a paragraph on p. 132, is the closest thing on the record to a full portrait of the man: "He was a physicist by night and an elevator engineer by day, and I never learned which of these he thought was his real work. He was, in either case, extraordinarily good at both."

A dev fact for the back pocket

killall does not mean the same thing on Linux and Solaris, and the difference has, over the last thirty years, permanently altered the career trajectory of an unknown but non-trivial number of system administrators.

On Linux — where killall is provided by the psmisc package written by Werner Almesberger at EPFL Lausanne in 1993 and shipped in essentially every Linux distribution since — killall <name> sends SIGTERM to every process whose executable name matches <name>. killall firefox closes Firefox. killall -9 chrome kills Chrome dead. The command has a -i flag that prompts for confirmation before each kill, a -u flag to restrict to processes owned by a specific user, and generally behaves like a pkill with a slightly different filter grammar.

On Solaris — where killall is /usr/sbin/killall, shipped as part of the base operating system since SunOS 5.0 (Solaris 2.0) in June 1992 — killall takes no argument. It sends SIGTERM to every process on the system except the caller, the caller's parent, and a small set of essential daemons. If any process is still alive after five seconds, it sends SIGKILL. It is called, on Solaris, by exactly one program: /etc/init.d/shutdown, during the transition from run-level 2 to run-level 0. It is the last thing that runs before the machine halts.

A Linux administrator sitting down at a Solaris console for the first time, in the reasonable belief that killall <name> will kill processes by name, has in fact requested that the operating system shut down every user process on the machine. On a workstation this is inconvenient. On a production database server it is, in the technical sense of the term, a resume-generating event.

The Solaris manual page has been warning about this since 1996, in the top-of-page CAUTION block on killall(1M) (still visible in the Solaris 11.4 documentation, section 1M, first paragraph): "killall is used by shutdown(1M) to kill all active processes not directly related to the shutdown procedure." The FreeBSD project, faced with the same conflict, resolved it in 1995 by adopting the Linux semantics for killall and providing the Solaris-style shutdown behaviour under the name killall5, matching the SysV convention. Linux's killall5, from the sysvinit-utils package, is the same command.

The lasting engineering lesson is not "check the man page" — everyone will, at some point, fail to check the man page — but that the same command name, in two closely related Unix branches, was independently given opposite semantics by teams who did not think to consult each other, and neither team was wrong given the context in which they were working. Solaris's killall was written for the shutdown script and named descriptively. Linux's killall was written for interactive use and named descriptively. The fact that "kill all" and "kill all matching" collapse to the same seven-letter binary is a coincidence of the shortness of English words in this domain. The command has, in effect, no globally agreed meaning. Any script portable across both systems must use pkill — which is unambiguous — or must check uname -s first and branch accordingly.

Primary sources:

• Werner Almesberger, original announcement of psmisc to comp.os.linux.announce, 7 September 1993, subject "psmisc: killall & pstree". Archived on Google Groups; also referenced in the psmisc-22.x source tree under THANKS.
• Oracle Corporation, Solaris 11.4 Reference Manual, section 1M, entry killall(1M). Available on-line under Solaris 11.4 man pages at docs.oracle.com.
• Sun Microsystems, SunOS 5.0 Administrator's Reference Manual, Sun Microsystems Inc., Mountain View, California, 1992, section 1M pp. 4-311 through 4-312. The original documented behaviour, unchanged in every subsequent Solaris release through Oracle Solaris 11.4.

If friend is ever administering a mixed shop, the answer is pkill. Not killall. pkill.

Today's goal

Delete one file you're keeping just in case.

Any file. The old draft. The screenshot from March. The final_v3_actual that was superseded by final_v3_actual_use_this that was superseded by the current version. The CSV export you told yourself you'd re-analyse. Pick one. Delete it.

The instinct that says "I might need this" is almost always right in the specific and almost always wrong in the aggregate. In the specific, yes, occasionally you will regret it. In the aggregate, the folder full of files you are keeping just in case is the reason you cannot find the file you actually need.

One file. Not the whole drive. That is a different day's exercise.

Today's toy in the corner is fleuron — a small SVG ornament generator. Each press produces one printer's flower, symmetric and useless in the strict sense of the word. A kickshaw, exactly.

Go build something small and out of proportion to its purpose.

— C