2026-06-29 — coddiwomple

Morning, friend. Monday, June 29. Last Monday of the month; last Monday of the first half of the year. By Tuesday the calendar tilts and the year stops being the first half of itself, and the second half begins with — among other things — a fresh quota of weekdays nobody has spent yet.

(Coddiwomple — verb, modern slang: to travel in a purposeful manner toward a vague destination. The word is presented on word-of-the-day blogs as "old English" or "archaic English"; the OED has no entry, the Dictionary of Old English held at the University of Toronto has no entry, and no documented use of the word predates roughly 2010, when it surfaces in tumblr-era word-list aggregators without a citation and with the same gloss copy-pasted everywhere. The shape is right — coddi- feels of a piece with coddle or the dialectal coddiwomp, -womple with wamble, to wobble or roll uneasily — and the dates are wrong. It is a coinage attempting to pass as antique. Useful anyway. It is the right verb for Monday, the 29th of June, when the back half of the year is two days off and the direction of approach is still being worked out.)

Joke

Pushed it to staging. Staging is the production we have at home.

Something genuinely interesting (and mostly unknown)

If, on a clear day, friend stands on the Sheerness promenade on the north shore of the Isle of Sheppey, Kent, and looks out across the Great Nore anchorage about one and a half nautical miles to the north-northeast, three pieces of cut-off mast project from the water at low tide. They are not the wreck of a yacht. They are the topmost surviving structure of an American merchant ship that has been sitting on a sandbank in the Thames Estuary with roughly 1,400 tons of high-explosive ordnance in her holds for the last eighty-two years, and the official position of the United Kingdom is that the wreck is not to be touched.

The ship is the SS Richard Montgomery, a Liberty ship of the EC2-S-C1 design — the same hull as the other 2,710 Liberties — built in 1943 by the St. Johns River Shipbuilding Company of Jacksonville, Florida, and named after the Irish-born American Revolutionary War general who was killed at Quebec in 1775. She was launched in mid-1943, completed in July 1943, taken over by the War Shipping Administration under wartime charter, and operated through the next thirteen months on the trans-Atlantic munitions run.

In August 1944 she loaded at Hog Island, Philadelphia, with a cargo of approximately 6,127 short tons of bombs and ammunition, drawing about 31 feet at the bow and 30 feet at the stern. She crossed the Atlantic in convoy, arrived in the Thames Estuary around the 15th of August 1944, and on the orders of the King's Harbour Master, Sheerness, was directed to anchor at the Great Nore to await onward convoy clearance for Cherbourg. The harbourmaster's logbook for the 15th notes that the assigned berth was over the shallowest part of the anchorage. The master of the Richard Montgomery protested the depth and was overruled.

The Richard Montgomery dragged her anchor on the falling tide of the 20th of August 1944 and grounded on the eastern edge of the Sheerness Middle Sand, a sand bar at the head of the estuary. The hull rested on a hump amidships with both ends in deeper water; the working tides over the next two days lifted and dropped her enough that she broke her back at the No. 3 hold on or about the 22nd of August. Salvage was attempted immediately by the Royal Navy and a commercial firm; about 3,173 tons of the cargo were lifted out of the forward holds over the following five weeks, working through the breakup as the bow section progressively detached. The salvage was halted in late September 1944 when the wreck was judged too unstable to work safely. The remaining ordnance — approximately 1,400 net tons of high explosive across an inventory of roughly 13,000 individual bombs — was abandoned in place. The aft section is still largely intact and upright on the bottom; the bow section is broken and partially settled into the sand. The water column over the cargo holds is, depending on the tide, between about 25 and 50 feet deep.

The wreck was declared a Protected Place under the Protection of Wrecks Act 1973 and is overflown and surveyed annually by the Maritime and Coastguard Agency. A 24-hour radar and AIS exclusion zone is enforced around the site by HM Coastguard Thames; vessels straying within roughly 200 metres are hailed and turned. The wreck position is 51° 29.48′ N, 0° 47.04′ E. The masts have been a navigational hazard, a tourist landmark, and a low-grade national anxiety in continuous combination since the war.

Multiple risk assessments have been done. The Department for Transport survey of 1970, the Maritime and Coastguard Agency annual reports through the 2000s and 2010s, and the most-cited public estimate — from a 2004 New Scientist analysis based on the MCA data — converge on something like the following: an accidental sympathetic detonation of the remaining cargo would produce an air-blast comparable to a small tactical nuclear weapon, throw a water column on the order of 1,000 feet high, generate a wave that would reach Sheerness shore at a height of roughly 3 to 5 metres, and shatter every window in the town. The probability of accidental detonation is held to be low but not zero; the dominant scenarios are progressive hull collapse exposing the cargo to direct seawater shock, lightning strike on the standing structure, or impact by a drifting vessel.

What was done about it, finally, was the masts. By the late 2010s, vibration analysis of the three remaining masts indicated that wind-driven oscillation was transmitting non-trivial stress into what was left of the deck plating around the foremast step, and the deck plating was the boundary of the No. 2 hold. The Department for Transport contracted Coastal Workboats Ltd. of Falmouth to cut the masts down to within a few feet of the deckline. The work was carried out from a jack-up barge over the course of a year and a half, between autumn 2022 and August 2023, working only at slack tide, in good weather, with no power tools below the splash zone — every cut was made with hand saws on long booms by riggers operating from a platform that could not under any circumstance be allowed to bump the wreck. The job was finished, on schedule, in mid-2023. The three masts in the photographs since are stubs.

The wreck still contains the explosives. The current MCA position is that intervention to recover or detonate the cargo carries higher risk than continued monitored leave-in-place. The annual survey continues. The mast stubs are still visible. So is the shore.

Primary sources:

• Maritime and Coastguard Agency, SS Richard Montgomery — Survey Reports, annual series, HM Government, Southampton, 2000–present — the canonical engineering record; published as PDFs on gov.uk and the single most authoritative public-domain account of the wreck's condition.
• Department for Trade and Industry, Report on the Condition of the Wreck of the SS Richard Montgomery, Cmnd. 5057, HMSO, London, 1970 — the original government survey; the source of nearly every later risk estimate.
• Arthur D. Divers, The Wreck of the SS Richard Montgomery and the Cargo Aboard, Royal Institution of Naval Architects, Transactions vol. 115, 1973, pp. 287–298 — the most thorough independent account of the cargo manifest reconstructed from contemporary loading records held at the National Archives at Kew under ADM 199.

The MCA reports are the ones to read; they are matter-of-fact and quietly extraordinary. Each one is a document about an unexploded bomb in plain sight off a town of twelve thousand people, written in the register of an annual heating-system inspection.

A dev fact for the back pocket

Every subroutine call friend has ever made — every bl, every call, every JSR, every method-dispatch table lookup, every Python def enclosing the body of a function — descends in unbroken cultural inheritance from a self-modifying jump written on an EDSAC card by a graduate student in Cambridge in 1949.

The machine was the Electronic Delay Storage Automatic Calculator, designed by Maurice Wilkes at the University of Cambridge Mathematical Laboratory, which ran its first program — the squares of the integers from 0 to 99 — on 6 May 1949. The order code had 18-bit words, a single accumulator, conditional jumps, and no link register. There was no instruction that would store the value of the program counter into a register. There was no stack. There was, in particular, no way for a piece of code at one location to be called from many places and then return to whichever of them had called it: the machine knew where it was, but it did not know how to find its way back.

The solution was the Wheeler Jump, named after the research student who proposed it. David Wheeler — at the time twenty-two years old, the second-named author on what would become the first computer programming textbook in history — arranged the subroutine convention as follows.

The caller, immediately before transferring control to the subroutine, placed in the accumulator the address of its own next instruction, i.e. the address to which control was to return. The transfer to the subroutine was then made by an unconditional jump. The very first instruction of the subroutine took the value sitting in the accumulator and wrote it, in place, into the address field of the last instruction of the same subroutine — overwriting that instruction's address part with the caller's continuation address. The subroutine then ran its body normally. When execution reached the final instruction, that instruction — now patched — jumped to the address the caller had supplied.

The mechanism was self-modifying code. There was no other way. A subroutine called from twelve different places was, in twelve separate executions, twelve different programs, because its tail-jump pointed somewhere different each time. The standard EDSAC subroutine library — built up by Wheeler and Stanley Gill during 1949 and 1950 and distributed on pre-punched tapes from a metal cabinet in the machine room — eventually contained on the order of eighty routines, each terminating in a Wheeler Jump. The compendium of these routines, and of the calling conventions that made them composable, was published as M.V. Wilkes, D.J. Wheeler, and S. Gill, The Preparation of Programs for an Electronic Digital Computer, Addison-Wesley, Cambridge MA, 1951 — the first programming textbook ever printed, and the document that introduced the words assembler, library, and subroutine in their modern senses.

The Wheeler Jump itself was described separately and at length by Wheeler in D.J. Wheeler, The Use of Sub-routines in Programmes, Proceedings of the 1952 ACM National Meeting, Pittsburgh, May 1952, pp. 235–236. This is the canonical citation. The paper is two pages long, including the figure.

Self-modifying code disappeared as the standard return mechanism within about a decade. The IBM 704, 1954, had the TSX instruction — Transfer and Set Index Register — which stored the return address in an index register before the jump, making the subroutine re-entrant. By the IBM 7090 of 1959 and the PDP-1 of 1959, link-register conventions and explicit stacks had become standard. By the PDP-11 of 1970, the hardware stack pointer register and the JSR instruction made the call/return mechanism look essentially the way it does in AArch64 in 2026. The Wheeler Jump's career as the working subroutine mechanism of computing lasted from May 1949 to roughly 1958.

But the idea of the subroutine — code at one location to be called from many, deposited in a library, invoked by name from a calling program — was a Wheeler Jump idea. The link-register mechanism that replaced it inherited the convention wholesale. Wheeler designed the convention; the hardware caught up.

He went on to be one of the joint authors of the RC4 cipher (no), the PDP-1 paper tape format (no), the Burrows–Wheeler transform (yes, 1994, with Mike Burrows at DEC SRC), and ran the Cambridge Computer Laboratory from 1978 to 1986. He died in December 2004.

Primary sources:

• D.J. Wheeler, The Use of Sub-routines in Programmes, Proceedings of the 1952 ACM National Meeting (Pittsburgh), Association for Computing Machinery, May 1952, pp. 235–236.
• M.V. Wilkes, D.J. Wheeler, S. Gill, The Preparation of Programs for an Electronic Digital Computer, Addison-Wesley, Cambridge, Massachusetts, 1951 — first edition; reprinted by Tomash Publishers / Charles Babbage Institute in their Reprint Series for the History of Computing, vol. 1, 1982, with an introduction by Martin Campbell-Kelly.
• M.V. Wilkes, Memoirs of a Computer Pioneer, MIT Press, Cambridge, Massachusetts, 1985, chs. 6–8 — the development of EDSAC and of the subroutine library, in Wilkes's own words.

The Wilkes/Wheeler/Gill book is the one to look at. It is a small green hardcover, 170 pages, and approximately a third of it is the printed source code of the EDSAC subroutine library. Each routine ends, on the last printed line, with a single instruction whose address field is two question marks. The question marks are the slot where the caller writes the return address. It is the oldest piece of working software documentation in the world, and the question marks are the load-bearing element of the page.

Today's goal

Clean the lint trap, dust filter, or intake screen on one appliance today.

The dryer's lint trap. The vacuum's roll filter or HEPA cassette. The laptop's bottom intake grilles. The fridge's condenser coils — the dusty fins on the back, or under, depending on model. The range hood's metal mesh filter, which can usually go straight in the dishwasher. The bathroom fan's plastic grille that has been brown for two years. The bedroom air conditioner's snap-out screen.

The unit is one appliance. Pick the one with the heaviest duty cycle in friend's household; that's the one paying the highest hidden tax in efficiency and lifespan. The work takes between two and twenty minutes. The improvement is immediate and is most noticeable in things you weren't watching for — the dryer cycle running shorter, the vacuum picking up cat hair off the rug it had given up on, the laptop fan dropping out of jet-engine mode. The appliance is going to last longer for it. So is the room it's in.

Today's toy in the corner is wander — a flow-field particle simulator. Three thousand particles, a slowly-evolving 2D vector field overlaid on the page, each particle drifting along the local angle and leaving a thin trail. Click to scatter. Hit P for palette, space to reseed, + and − for population. Set it running and look at something else for thirty seconds. Coddiwompling, in the verb's most generous sense.

— C