2026-08-28 — snollygoster

Morning, friend. Friday, and the last one in August — the calendar is about to turn on the hinge it always turns on louder than the days on either side of it. The cottonwoods on the corner have gone from "you'd notice if you looked" to "you notice without looking." The mornings are pulling on a sweater and the afternoons still aren't. Somebody down the street is running a leaf blower against three leaves.

(Snollygoster — American English noun, informal and derisive, chiefly nineteenth-century political; a shrewd, unprincipled person, especially a politician who conducts themself entirely for personal advantage under the outward forms of public service. Origin uncertain; the standard reconstruction, defended in Mitford M. Mathews's A Dictionary of Americanisms on Historical Principles (University of Chicago Press, 1951) and reprinted in every subsequent American slang lexicon of the twentieth century, traces the word to the Pennsylvania German schnelle Geeschter — roughly, "quick spirits," a folk-name for a will-o'-the-wisp — carried westward from the print culture of the Ohio Valley in the eighteen-forties. The earliest attested American use in print is a paragraph in the Columbus weekly newspaper The Ohio State Journal on 28 October 1846, describing an unnamed county-level Whig officeholder as "a very slippery snollygoster." Harry S. Truman, at a routine press conference at the Key West Naval Base in October 1952, used the word publicly and was pressed by White House correspondents for a definition. He replied — in the general reconstruction his contemporaries recorded and Merriam-Webster now cites — that a snollygoster was a man in office guided by neither principle, party, nor country, but only by the desire for the office itself and the small perquisites of the office. Truman told the reporters to print it. They did. The word is one of the very small number of American political epithets — with bunkum, filibuster, gerrymander, lobbyist, pork barrel — that entered the lexicon from local politics and has stayed useful ever since. Every era has its snollygosters, and every era discovers, with the small unpleasant surprise of the freshly-minted, that the word already existed.)

Joke

The keynote demo is a video of the demo.

Something genuinely interesting (and mostly unknown)

On the afternoon of Sunday, 10 November 1872, in the private offices of the Bank of California at the corner of California and Sansome Streets in San Francisco, the bank's president William Chapman Ralston — the man whose personal signature had, more or less, financed the settlement of the American West Coast for the preceding decade — signed a cheque for six hundred sixty thousand dollars payable to two Kentucky prospectors named Philip Arnold and John Slack in exchange for a two-thirds interest in a diamond-and-ruby claim on a mesa in the Uinta country straddling the Colorado–Wyoming line. The remaining third had been placed, on Ralston's insistence and against Arnold's carefully-performed reluctance, in a trust controlled by a syndicate of some twenty-five investors, capitalised at ten million dollars, whose officers included the former Union Army general George S. Dodge and the mining engineer Asbury Harpending, and whose credibility rested on the appraisal — delivered in New York in October — of a jeweller named Charles L. Tiffany, of Tiffany and Company, who had personally examined an unmounted parcel of the raw stones and had valued them at approximately one hundred fifty thousand dollars for that parcel alone, extrapolating from which the syndicate's mining engineer projected a claim value into the tens of millions. The mesa contained no diamonds and no rubies. Arnold and Slack had, over the preceding fourteen months, spent something like thirty-five thousand dollars in cash acquiring rough industrial-grade diamond and ruby stock from cutters in London, Amsterdam, and Frankfurt, transported the stones across the Atlantic and overland to the Uinta country in personal luggage, and scattered them by hand across a specific quarter-acre of high desert north of the Colorado line. The exposure came three weeks later, in the second week of November, when a young federal geologist named Clarence King — director of the U.S. Geological Exploration of the Fortieth Parallel and, at thirty, one of the more accomplished field geologists then working in America — rode out to the site with three assistants and a small camp, and, within two days, established from a systematic mapping of the salted zone that the stones occurred in a distribution no natural process could have produced.

The tells were geological rather than mineralogical. King's party — Allen D. Wilson, Samuel Franklin Emmons, and A. R. Marvine, all trained field men — noticed on the second morning that diamonds and rubies co-occurred on the same square yard of ground, an association impossible in nature (the two form in different rock types at different depths, and no known geology produces them together); that the stones were confined precisely to the area an adult on foot could reasonably scatter to, some quarter-acre; that no stone occurred more than four inches below the surface, and only in soil recently disturbed by hand or boot; that the ratios of diamonds to rubies across the site were, to a degree of statistical peculiarity, identical to the ratios in the samples Tiffany had appraised in October; and — the detail King later published — that in the middle of the claim was a specific ant hill from which a small ruby had been extracted by one of the syndicate's own witnesses on an earlier visit, but which, on King's re-examination with a shovel, contained no additional rubies at any depth, and whose surface stones showed clear thumb-impressions in the loose earth around them. King telegraphed his conclusions to Ralston at the Bank of California on 11 November 1872 and delivered his full report to the syndicate's board four days later.

Arnold and Slack had by then already left San Francisco. Arnold, on being confronted, agreed to refund a portion of his fee in exchange for a promise of no prosecution, kept most of the money, returned to Elizabethtown, Kentucky, opened a small bank of his own, and died in 1878 of pneumonia contracted after a shootout with a rival Kentucky banker in the street outside his own front door. Slack, the more taciturn of the two, took his roughly fifty thousand dollars, moved to White Oaks in the New Mexico Territory, opened a coffin-making business, worked at it for twenty-four years, and died in 1896 without ever having spoken publicly about the affair. Ralston did not survive the aftermath by long. The Bank of California collapsed on 26 August 1875 — the diamond fraud was not the direct cause but was the founding event in a three-year erosion of investor confidence in Ralston personally — and Ralston himself was found dead in San Francisco Bay the following afternoon after swimming out from North Beach against a tide he had swum daily for years. The coroner's finding was drowning. The manner was left undetermined and has been argued about ever since.

Clarence King, on the other hand, was appointed the first director of the newly-consolidated United States Geological Survey in 1879 on the strength of, more or less, a fortnight of fieldwork on a Uinta mesa. The USGS in its current form is his direct descendant. And the exposure of the Great Diamond Hoax is the founding data point for a specific class of American swindle — the artificially-salted natural-resource claim, offered to outside capital on the strength of a third-party appraisal — every subsequent version of which has been recognised for what it was by geologists applying essentially the same class of statistical test King used on the second morning of November 1872, up to and including the Bre-X Minerals gold-salting fraud on the Busang concession in East Kalimantan in the mid-1990s, which was exposed in April 1997 when Freeport-McMoRan's independent drill program returned barren cores from holes twinned with the salted ones. What the salter can control is the stones on the surface. What the salter cannot control is the distribution.

Primary sources:

• King, Clarence. Report to the Directors of the San Francisco and New York Mining and Commercial Company. San Francisco, 25 November 1872. Reprinted verbatim as Appendix B in Woodard, cited below. The founding field report, drafted from notes taken between 5 and 11 November, containing the specific salting evidence.
• Harpending, Asbury. The Great Diamond Hoax and Other Stirring Incidents in the Life of Asbury Harpending. Edited by James H. Wilkins. San Francisco: The James H. Barry Company, 1913. The memoir of the syndicate's own mining engineer, published forty-one years after the fact when Harpending was the last surviving principal. His account of the London gem purchases, the New York Tiffany appraisal, and the November confrontation with Arnold is the primary insider record.
• Woodard, Bruce A. Diamonds in the Salt. Boulder, CO: Pruett Press, 1967. The standard modern monograph on the affair, drawing on the surviving Bank of California records in the Wells Fargo History Room, the Arnold family papers in the Filson Historical Society in Louisville, and the King papers at the Huntington Library. Includes the reconstructed itinerary of the 1871 European gem-buying trip.
• Wilkins, Thurman. Clarence King: A Biography. Revised and enlarged edition. Albuquerque: University of New Mexico Press, 1988. The definitive treatment of the field investigation. Includes the King party's daily field notes for 5–11 November 1872, transcribed from the original notebooks at the Huntington.

A dev fact for the back pocket

On every x86-family personal computer shipped between the introduction of the IBM PC AT in August 1984 and, in most consumer chipsets, some point between 2008 and 2012, exactly one physical address line — the twenty-first, counting from zero, formally address bit A20 — was routed through an external gate whose control input was a single output bit of the Intel 8042 keyboard controller. The purpose of the gate was to enforce, at will, a specific address-arithmetic behavior that had been an accidental property of the 8086 processor of 1978 and that a small but very widely-shipped body of 8086 software had come to depend on: that a segment-plus-offset computation exceeding the 20-bit range of the 8086's external address bus would silently wrap back to physical address zero, so that the theoretical address FFFF:0010 (segment plus offset equal to 100000h, which is one byte past the first megabyte) resolved on real 8086 hardware to physical address 00000000h. The 80286 of 1982, with a 24-bit address bus, no longer wrapped: the same computation now addressed real memory a fraction of a kilobyte above the megabyte line. The software that depended on the wrap — the CP/M-86 program-entry convention CALL 5, several DOS internal calls, and a handful of contemporary application programs including early releases of Lotus 1-2-3 and WordStar — would silently corrupt data in the wrong region of memory. IBM's fix, shipped in the AT ROM BIOS and reproduced by every subsequent PC-clone manufacturer, was to install an external AND gate on the physical A20 line between the CPU and the memory bus and to route the enable input of that gate to a general-purpose output bit of the 8042 keyboard microcontroller, so that DOS could, by writing a byte to the keyboard port, choose whether address bit A20 was to be forced low (wrap, for compatibility) or passed through (full addressability, for anything above the 1 MB line). This design survived every subsequent x86 generation for approximately twenty-five years.

The specific mechanism repays walking through. On the AT motherboard, physical address line A20 emerged from pin A1 of the 80286 socket adjacent to the 8042. Between that pin and the memory bus, IBM's engineers placed a single gate of a 74LS08 quad AND package, one of whose four gates had inputs wired to (a) the CPU's A20 output and (b) a signal called A20GATE driven from the 8042's output port bit 1. The 8042 firmware, running on an Intel 8048 embedded microcontroller at approximately 6 MHz, provided two paths for the operating system to raise and lower A20GATE: command D1h ("Write Output Port") followed by a data byte whose bit 1 controlled the line, and — the specifically bizarre part, added because the D1h path was inconveniently slow — dedicated single-command shortcuts DDh ("Disable A20") and DFh ("Enable A20"). Because the 8042 handled these commands on the timescale of its own microcode — typically several dozen microseconds per command, glacial by CPU standards — operating systems that needed to toggle A20 rapidly (Windows 3.0 in 386 enhanced mode; the OS/2 kernel; the early Linux kernel) resorted to hardware-specific fast paths that bypassed the 8042 altogether. The most widely-supported of these was the PS/2 System Control Port at I/O address 92h, whose bit 1 was wired in parallel with the 8042 output on nearly every clone chipset shipped after 1987, and which allowed the OS to toggle A20 with a single OUT instruction rather than waiting for a keyboard-controller round-trip.

The A20 gate was documented in the IBM Personal Computer AT Technical Reference (IBM Publication 1502494, first edition March 1984) in the "System Board Description" chapter, with a footnote noting only that "the A20 GATE signal exists to preserve compatibility with software written for the IBM Personal Computer and Personal Computer XT." No further explanation was offered; none was needed. Every clone-maker copied it. HIMEM.SYS from MS-DOS 5.0 (June 1991) used the A20 gate to expose the High Memory Area — a 65,520-byte region from physical 100000h to 10FFEFh, accessible only when A20 was enabled and addressable in real mode as segment FFFFh — recovering the last usable slab of memory beneath the second megabyte on machines that had a full megabyte-plus of DRAM installed. One gate, on one line, driven by one bit of a keyboard controller, was for the fifteen years between 1985 and 2000 a load-bearing component of the DOS memory manager on every IBM-compatible PC on earth.

The gate did not die until the ninth generation of x86 CPUs. AMD dropped support for the A20-mask hardware pathway from its K10 family in 2007; Intel followed with the Nehalem-based Core i7-900 series in November 2008, and by the time Ivy Bridge shipped in 2012 the mask was gone from the client line entirely. But by then a decade of BIOS, bootloader, and early-kernel code had been written against the assumption that A20 needed enabling; the ceremony was harmless and the code was in place, so it stayed. In 2026 every x86 desktop and laptop boots through code that pretends to talk to a 1984 keyboard controller for approximately three microseconds on its way into protected mode. The pretense is performed, on every Linux boot, by the function enable_a20() in arch/x86/boot/a20.c, which tries four methods in sequence — BIOS interrupt 15h/AH=24h, port 92h, the 8042 command DFh, and a memory-wraparound test — and returns as soon as any of them appears to have worked. The keyboard controller is not there. The gate is not there. The convention is there because it was always there.

Primary sources:

• International Business Machines Corporation. IBM Personal Computer AT Technical Reference. First edition, IBM Publication 1502494. Boca Raton, FL: IBM Corporation, March 1984. The "System Board" chapter contains the schematic and functional description of the A20GATE signal and its routing through the 8042 keyboard controller.
• Intel Corporation. iAPX 286 Hardware Reference Manual. Order Number 210760-001. Santa Clara, CA: Intel Corporation, 1983. Chapter 3, "CPU Pin Function Description," covers the physical address lines A0 through A23. The manual does not itself specify the A20 gate — the gate was IBM's addition to the AT motherboard, not Intel's — but its description of the address bus is the reference for what the gate is intercepting.
• Microsoft Corporation. MS-DOS 5.0 Programmer's Reference. Redmond, WA: Microsoft Press, 1991. Chapter 15, "Extended Memory," documents HIMEM.SYS and the eXtended Memory Specification (XMS) call sequence. The A20 gate is treated on pages 481–486 as the specific mechanism by which HIMEM.SYS reaches the High Memory Area.
• Collins, Robert R. "The Segment:Offset Wraparound." Dr. Dobb's Journal, vol. 20, no. 9, September 1995, pp. 44–49. The definitive published summary of the A20 gate's history through the mid-1990s, including the eight distinct methods by which contemporary operating systems toggled it (via 8042, via port 92h, via BIOS INT 15h AH=24h, and five vendor-specific chipset paths) and the compatibility considerations that had kept the gate in silicon through three generations of x86 processors up to that point.

Today's goal

Notice one moment today when friend is about to shade the truth for small personal convenience — the polite lie, the hedged commitment, the "I'll get to it" both parties know is a soft no, the compliment that isn't meant, the answer that's technically true and functionally misleading — and, in that specific moment, say the flat true thing instead. One moment. Not a program. Not a personality reform. One line, one time, where the small easy false thing was queued up and ready and the small harder true thing comes out instead.

The snollygoster is not a monster. He is the small daily version of the same thing everybody does. He is a person who noticed, at some point, that the friction of the true answer was greater than the friction of the plausible-sounding false one, and who has, over years, made the substitution more or less by reflex — on the small stuff, at first, then progressively on the larger. He is downstream of a great many individual moments in each of which a small easier false thing was said instead of a small harder true thing, and in none of which was anything of consequence lost. That is the arithmetic. Any single moment of it is negligible. The sum, at a working life's scale, is not.

The exercise is the same one in reverse. One moment, one line, one true thing where the easy false one was already forming. The moment is not flagged in advance; you have to catch it in flight. That catch is most of the exercise. The saying — the actual small sentence containing the true thing — is almost always less costly than the anticipation of it suggests. The other person, in the great majority of cases, does not react at all. The friction cost is small. The alternative — the sum of a thousand of these small unresolved substitutions — is not.

One moment. Say the true thing. Move on.

Today's toy is the stump — a small nineteenth-century American whistlestop platform, in a browser tab, that produces one complete political stump speech from a period-flavoured snollygoster every time you pull the lever. Each speech comes with a candidate, an office being sought (register of deeds, canal commissioner, county coroner, town supervisor, prothonotary), a party (Whig, Free Soil, Know-Nothing, Anti-Monopoly, Greenback, Prohibition), a year between 1840 and 1900, a town, an opponent never named directly, and three paragraphs of confident and grammatically impeccable nonsense. Pin the good ones to the shelf; the browser remembers them. Lives in the corner.

Go say the true thing, friend.

— C