2026-08-10 — collieshangie

Morning, friend. Monday. First working day of the week — the day the calendar is broadly optimistic on friend's behalf and the inbox has taken the weekend to arrange itself into a particular kind of small mess. Take the first hour slowly. The week will be there in the second one.

(Collieshangie — Scots noun, "a noisy or confused dispute; a hubbub; a squabble on the point of getting out of hand." Attested in Scots print by the late eighteenth century and common in Border English through the nineteenth. Sir Walter Scott has it in Guy Mannering (1815): "there was sic a collieshangie kick'd up i' the kirkyaird." The etymology is itself a small collieshangie. The Scottish National Dictionary favours colley — the Scots working sheepdog of the period, before the word narrowed to the modern collie breed — plus a syllable of loose intensifying compound; a collieshangie is what happens when two dogs on the same street get into it. A minority reading, going back to John Jamieson's Etymological Dictionary of the Scottish Language (1808), reads the second half as shangie, a link or short chain, particularly the chain used to fasten a dog to a post: a collieshangie is what happens when two chained dogs reach the ends of their tethers at the same time. Both accounts point at the same picture. The word declines in ordinary written English after about 1900 and now survives largely in Doric Scots. Monday is when the week rediscovers who else lives on the block.)

Joke

A microservices migration is a monolith having a collieshangie with itself, in three availability zones, on separately-invoiced infrastructure.

Something genuinely interesting (and mostly unknown)

At 09:31:02 Pacific time on 5 April 1958, on the west coast of Vancouver Island, the Canadian government detonated 1,375 tonnes of a proprietary aluminized ammonium-nitrate slurry called Nitramex 2H inside a mountain that was underwater. The mountain had been climbed to from below, through a shaft sunk on a nearby island and a horizontal tunnel driven three-quarters of a kilometre out under the seabed. The point of the operation was to lower the mountain's twin summits by about twelve metres so that ships could cross over them. The explosion was, at the time, the largest planned non-nuclear detonation in history. It was broadcast live on national television.

The mountain was Ripple Rock, a twin-peaked underwater pinnacle rising off the floor of Seymour Narrows, a two-kilometre choke point on the Inside Passage between Vancouver Island and Sonora Island in coastal British Columbia. The Narrows carries tidal currents that reach 15 knots on the flood — among the fastest tidal streams in navigable water anywhere in the world — and Ripple Rock sat exactly in the middle of the channel. At low tide the peaks stood 2.7 metres below the surface; at low low water in the spring tides they broke the surface entirely. Captain George Vancouver in 1792 called the passage "one of the vilest stretches of water in the world" in the log of HMS Discovery and gave the Narrows a wide berth. Between the arrival of steam navigation in the 1870s and 1958, at least 20 large vessels and 100 or so smaller craft were wrecked on the rock, at a documented cost of 114 lives.

The Canadian government considered doing something about it for fifty years. Proposals from 1931 onward included building a suspension bridge over the Narrows using Ripple Rock as the central pier; running a lighthouse and a foghorn on a caisson pinned to the rock; and blowing the peaks off with charges placed by divers on the surface. The bridge and lighthouse were ruled out on the current. The dive-placed charges were tried three times between 1943 and 1945 — the technique was to moor a floating barge above the peaks, drill down into the rock through drill strings suspended from the barge, and lower explosives in cased charges — and abandoned in 1945 after nine workers were killed in various small accidents by cables parting under tidal load. The current tore the barge moorings out roughly once a month.

The 1953 solution, designed by the consulting engineers Dolmage & Mason of Vancouver under contract to the Department of Public Works, was to attack the rock from below the seabed. A vertical shaft would be sunk on Maud Island — a small island 200 metres from the rock on the Vancouver-Island side — to a depth well below the seabed. From the bottom of that shaft, a horizontal drift would run out under the Narrows to the base of the rock. Two upward raises would then be driven inside the rock itself, hollowing out the two peaks from the inside like the inside of a tooth. The chambers would be loaded with explosive; the tunnel would be backfilled with concrete; the surface would be evacuated; the charge would be fired.

Construction ran from November 1955 to 5 April 1958. The Maud Island shaft went down 174 metres. The horizontal tunnel ran 762 metres out under the seabed at a slight downward pitch. The upward raises into the two peaks were 91 metres and 107 metres respectively. The tunnels were driven by a crew of about 75 miners, working three shifts, using a compressed-air drill and pneumatic mucking cars. The tunnel intersected several unmapped faults carrying sea water, all of which had to be grouted and re-sealed on the way through; the concentration of sea water in the drift air had to be monitored continuously because the compressor intakes were shore-side and any leak would have been detected first by the drills seizing on rust. Three miners died over the course of the operation, one in a rockfall and two in a tramway accident on Maud Island.

The explosive was 1,375 tonnes of Nitramex 2H, an aluminized ammonium-nitrate slurry developed by Du Pont Canada specifically for the job. The charge was distributed across 2,756 loading holes drilled outward from the two upward raises into the interior of each peak. Firing was in a staged sequence over about 2 seconds to shape the outward push and keep the shock inside the rock rather than propagating along the tunnel back toward Maud Island. The tunnel and shaft were sealed with 300 metres of concrete plug before firing.

The Canadian Broadcasting Corporation ran a live outside broadcast, produced from a portable control room at the Maud Island headworks, fed by microwave hop across the Strait to the Vancouver studios and from there through the CBC's national relay. The audience was estimated at 1.2 million — roughly one in fifteen Canadians alive at the time. The detonation lifted a column of water and pulverised rock estimated at 370,000 tonnes to a height of about 300 metres. The camera survived; the microwave link held; the announcer, Bill Herbert, kept talking through the shock wave, which reached the Maud Island shore approximately 12 seconds after ignition. The Narrows briefly went dry over the crater as the water was displaced, then reflooded within half a minute.

The peaks were reduced to at least 13.7 metres below chart datum. Ripple Rock is now marked by a fathometer trace and a note in the Canadian Hydrographic Service Chart 3543 to the effect that mariners transiting Seymour Narrows should time their transit to slack water and give the site half a cable of clearance. It has caused no reported wrecks since. The tunnels beneath Maud Island are sealed and the shaft is capped; a Parks Canada interpretive plaque on the shoreline gives the coordinates and the tonnage.

Primary sources:

• Dolmage, Victor and Mason, John T. "The Removal of Ripple Rock." Transactions of the Canadian Institute of Mining and Metallurgy, vol. 62, 1959, pp. 41–56. The as-built engineering report by the consulting engineers who designed the operation. Includes the drift alignment, the charge-hole pattern, and the firing sequence.
• Department of Public Works of Canada. "Ripple Rock: The Story of a Great Engineering Achievement." Ottawa, 1958. The official DPW pamphlet, released the week of the detonation, with the operational tonnages, timings, and personnel roster. Held at Library and Archives Canada, RG 11 series.
• Turnbull, Elsie G. Ripple Rock: The Rock That Wouldn't Die. Heritage House, Surrey, BC, 1974. The narrative history, drawing on the Maud Island crew logs, DPW files, and interviews with the surviving Dolmage & Mason engineers and the CBC broadcast crew.

A dev fact for the back pocket

The Windows API represents time as 100-nanosecond intervals since 00:00:00 UTC on 1 January 1601. Both parts of that specification — the tick size and the epoch — were engineering decisions with histories. Neither of them originated at Microsoft.

The API is the FILETIME structure, present since Windows NT 3.1 in July 1993 and unchanged through every subsequent 32- and 64-bit Windows. It stores a signed 64-bit integer of ticks, where one tick is 100 nanoseconds. A FILETIME of 0 represents the first tick after midnight on 1 January 1601 UTC. A FILETIME at the far end of its range can represent times up to approximately 30828 AD.

The tick size was inherited from OpenVMS, where the system time is represented as 100-ns ticks since 00:00:00 UT on 17 November 1858. VMS's epoch is the Modified Julian Date epoch, chosen in the 1950s by the Smithsonian Astrophysical Observatory's satellite-tracking programme so that the current-day Julian date would fit into 18 bits — the word size of the observatory's tabulating machines. VMS's own epoch document — the VAX/VMS System Services Reference Manual, section on $NUMTIM and $BINTIM — is explicit about the provenance and refers the reader to the Smithsonian tables. David Cutler and the VMS group at DEC through the 1970s used this format throughout the operating system; when Cutler was hired by Microsoft in 1988 to build Windows NT, he brought the tick-based time representation with him almost verbatim.

The epoch shift — from 1858 to 1601 — was NT-side, and its justification is calendar-arithmetic. The Gregorian calendar repeats on a 400-year cycle: every year divisible by 4 is a leap year, except centuries not divisible by 400, and after four centuries the sequence of leap and common years is bit-identical to the previous four centuries. 1 January 1601 is the start of the first Gregorian 400-year cycle whose end (2000-12-31) falls within any plausible Windows deployment horizon. Choosing that epoch means the routine that converts a FILETIME to a calendar date can subtract 1601, take the result modulo 400, and index into a precomputed table of exactly 400 years' worth of days — no special-casing century-year leap logic, no branch on which cycle the input falls in, no arithmetic on the input beyond a subtraction and a modulo. The equivalent Unix mktime code is several times as long because the Unix epoch (1970-01-01) sits in the middle of a Gregorian cycle rather than at its boundary, and every conversion has to reason about which cycle the input belongs to.

The canonical Microsoft explanation of the 1601 choice is a blog post by Raymond Chen, principal software engineer at Microsoft, on 6 March 2009: "The Windows time-of-day format is like the FILETIME format... The FILETIME epoch of January 1, 1601 was chosen because it's the beginning of the current 400-year cycle of the Gregorian calendar." Chen's blog, The Old New Thing, is the closest thing Windows has to an in-tree history file; the FILETIME entry has stood for seventeen years unchallenged.

The historical note that gets the most gasps: FILETIME was designed with a range extending to the year 30828 AD because Cutler wanted to be sure the OS's time representation would outlast any conceivable file it stored. Compared to time_t on 32-bit Unix, which rolls over on 19 January 2038, this is a design margin of roughly a factor of 9,300. The 32-bit Unix roll-over is currently a documented and funded engineering programme across most of the industry. The Windows roll-over is scheduled for a Wednesday.

Primary sources:

• Chen, Raymond. "Why is the Win32 epoch January 1, 1601?" The Old New Thing, MSDN Blogs, 6 March 2009. Archived at devblogs.microsoft.com/oldnewthing/20090306-00. The canonical Microsoft explanation.
• Digital Equipment Corporation. VAX/VMS System Services Reference Manual, various editions, 1978 onward. Section on $NUMTIM and $BINTIM. Documents the 17 November 1858 epoch, the 100-nanosecond tick, and the Smithsonian provenance.
• Custer, Helen. Inside Windows NT. Microsoft Press, Redmond, Washington, 1993. Chapter on the executive object manager and time services, written from inside the Cutler group; describes the design intent behind the FILETIME range and the deliberate inheritance from VMS.

Today's goal

Before friend opens any inbox, any feed, any message thread, or any news headline this morning, step outside for two minutes. Not a walk. Two minutes at whatever exit is nearest — a doorstep, a stoop, a balcony, the mouth of an alley. Look at the sky. Notice what the air is doing. Then go back in and start the week.

The purpose is not fresh air and it is not mindfulness. The purpose is to let the day arrive on friend's own schedule before anyone else's calendar tells friend what day it is. The first thing that touches a Monday morning tends to set the register for the rest of the day; if the first thing is a Slack notification, the day starts as a Slack day; if it is two minutes of ordinary weather that friend did not have to answer, the day starts as friend's day and then becomes a Slack day at roughly 09:15, on friend's terms.

Today's toy is collieshangie — a quiet field of small figures wandering peacefully in their own colours. Click anywhere and drop a squabble — a local disturbance that scatters neighbours, mixes the parties briefly, and dies down. Watch the field come back to peace. A slider sets the field's temperament: how long it holds a grudge. Lives in the corner.

— C