2026-07-28 — borborygmus

Morning, friend. Tuesday. The plan friend made Sunday night has already been renegotiated twice and it isn't lunch yet.

(Borborygmus — noun, from the Greek βορβορυγμός, the sound of gas moving through the intestines. Onomatopoeic — the word imitates the sound, which is unusual for a term that survived intact through Latin. Attested in Hippocrates around 400 BC in the context of the noises made by the digestive tract during illness, and used continuously through Galen and into the Byzantine physicians. It entered English in the early 18th century via translations of Thomas Sydenham's clinical notes and is still the standard clinical term in 2026 — the GP will use exactly this word, without irony, and write it on a chart. The plural is borborygmi. The sound friend's gut made in the standup this morning was one of them.)

Joke

The best documentation in any codebase is the git blame column, and it is always older than friend feared.

Something genuinely interesting (and mostly unknown)

At 20:32 UTC on 18 November 1929, a magnitude 7.2 earthquake broke on the continental slope about 250 km south of Newfoundland, in the area of the Laurentian Fan. Onshore, on the Burin Peninsula, twenty-eight people were killed, mostly by a 3–8 m tsunami that struck about two and a half hours after the tremor. The tsunami is the reason the earthquake is in local memory. It is not the reason anyone still writes about it.

Twelve transatlantic telegraph cables ran across the region — the North American end of the network that had been steadily thickened since the Great Eastern laid the first successful transatlantic cable in 1866. All twelve cables broke. But they did not break at the same time.

The five cables closest to the epicenter broke at 20:32 UTC, simultaneously, at the moment of the earthquake. The rest broke in sequence — each one further south, each one later than the last. The final cable failed at 09:49 UTC on 19 November, thirteen hours and seventeen minutes after the earthquake, roughly 800 km from the origin.

The cable companies — Western Union, the Commercial Cable Company, Anglo-American — kept precise operational logs of cable faults. A break cost money and required a repair ship to locate and splice the damaged section. Each fault was timestamped to the second, its position estimated from the resistance-to-fault measurement taken from both landing stations. The cable operators kept these records for insurance and billing. They were not, in any conscious sense, running a scientific instrument.

In 1949, Bruce Charles Heezen — twenty-four years old, working on his doctorate at Lamont Geological Observatory of Columbia University — noticed the cable-break sequence in the operator archives while researching submarine geology of the Grand Banks. He and his advisor Maurice Ewing worked backward from the break times. If a single event — an underwater avalanche of sediment, a turbidity current — had moved downslope from the epicenter, the break times gave you its speed. The math was arithmetic: distance divided by time. The answer was that the slide had been moving at about 55 knots (100 km/h) as it left the continental slope and had decelerated to about 12 knots (22 km/h) by the time it broke the last cable 800 km downslope. The estimated volume of displaced sediment was on the order of 200 km³.

The paper appeared in 1952: Heezen and Ewing, "Turbidity currents and submarine slumps, and the 1929 Grand Banks earthquake", American Journal of Science vol. 250, pp. 849–873. Before this paper, turbidity currents were a laboratory curiosity — small density flows PhD students produced in flumes. After it, they were a first-order geological force, the primary mechanism by which the continental slopes deliver sediment to the deep sea, and the reason the abyssal plains of the world's oceans are as flat as they are. The paper reshaped sedimentology from a set of cable-repair invoices.

The Western Union cable-fault ledger for the night of 18 November 1929 sits in a folder in the AT&T Archives, in Warren, New Jersey. It is a business record. It became a scientific dataset the moment somebody thought to read it that way.

Primary sources:

Bruce C. Heezen and Maurice Ewing, "Turbidity currents and submarine slumps, and the 1929 Grand Banks earthquake", American Journal of Science, vol. 250, no. 12, December 1952, pp. 849–873. The foundational paper. The cable-break timing table on pp. 858–860 is the entire argument.
Alan Ruffman, Tsunami runup mapping as an emergency-preparedness planning tool: the 1929 tsunami in St. Lawrence, Newfoundland, Geomarine Associates report to Emergency Preparedness Canada, 1996, EPC file 4110-49-1, three volumes. The definitive reconstruction of the Burin tsunami; contains transcriptions of the cable operators' original fault logs from the AT&T (formerly Western Union) archives.
David J. W. Piper, A. N. Shor, and J. E. Hughes Clarke, "The 1929 Grand Banks earthquake, slump, and turbidity current", in Sedimentologic Consequences of Convulsive Geologic Events, Geological Society of America Special Paper 229, 1988, pp. 77–92. Reanalyzes Heezen and Ewing with modern bathymetric survey data. The speed and volume estimates hold up.

A dev fact for the back pocket

The minimum Ethernet frame is 64 bytes, and the minimum payload is 46 bytes, because of a physical constraint on a variety of coaxial cable that stopped being installed in 1990. Every packet on every LAN in 2026 still pays the tax.

Robert Metcalfe wrote the original memo describing "Ether" — a shared broadcast medium — at Xerox PARC on 22 May 1973. Metcalfe and David Boggs published the first academic paper — "Ethernet: Distributed Packet Switching for Local Computer Networks" — in Communications of the ACM 19(7), July 1976, pp. 395–404. The DIX v1.0 standard (Digital Equipment Corporation, Intel, and Xerox) followed on 30 September 1980. IEEE 802.3 was ratified on 24 June 1983 and published as IEEE Std 802.3-1985.

The medium was 10BASE-5 — thick coaxial cable, half-inch outside diameter, yellow jacket, up to 500 m per segment, up to five segments through four repeaters for a maximum end-to-end network diameter of 2,500 m. Every station shared the cable. The access rule was CSMA/CD: listen before transmitting, and if two stations transmitted at once, both would detect the collision, send a jam signal, and back off.

For CSMA/CD to work, a transmitter has to be still transmitting when its own collision reaches it. The maximum round-trip time on a 2,500 m 10BASE-5 network — propagation delay plus repeater delay plus a margin — is 51.2 μs. At 10 Mbps, that is 512 bits. So the minimum frame, on the wire, is 512 bits, or 64 bytes: 14 bytes of header (destination MAC, source MAC, EtherType), 4 bytes of frame check sequence, and 46 bytes of payload. Anything shorter is a "runt", and a runt is presumed to be a collision fragment. If a station has less than 46 bytes to send, the driver pads.

The physical constraint stopped existing when the network became switched. 10BASE-T (IEEE 802.3i, 1990) put every station on a dedicated point-to-point link to a hub, and eventually a switch, with no shared medium and no collisions. 100BASE-TX (1995) reduced the slot time to 5.12 μs and pulled the maximum half-duplex diameter down to about 205 m to keep the 64-byte minimum working. Gigabit Ethernet (802.3z, 1998) at the same diameter would have required a 512-byte minimum frame — which would have broken every existing implementation — so IEEE invented "carrier extension": the MAC hands a 64-byte frame to the PHY, and the PHY pads on the wire out to 512 bytes with special non-data symbols. This wasted bandwidth on small frames. But the frame stayed 64 bytes. 10 Gigabit Ethernet (802.3ae, 2002) dropped CSMA/CD entirely — switched-only, full-duplex only — and kept the 64-byte minimum. It kept the 46-byte minimum payload.

An ARP request has 28 bytes of payload. On every LAN in the world in 2026, an ARP request is transmitted as 46 bytes, of which 18 are zero padding, because of a 1980 decision about how long a signal takes to travel down 2,500 metres of yellow coaxial cable that nobody has installed in thirty-six years.

Primary sources:

Robert M. Metcalfe and David R. Boggs, "Ethernet: Distributed Packet Switching for Local Computer Networks", Communications of the ACM, vol. 19, no. 7, July 1976, pp. 395–404. Section 3.3 (Interference Detection) is where the round-trip-time argument for a minimum frame length is first written down.
IEEE Std 802.3-1985, "IEEE Standards for Local Area Networks: Carrier Sense Multiple Access with Collision Detection (CSMA/CD) Access Method and Physical Layer Specifications", IEEE, ratified June 1983, published 1985. Clause 4.2.3.3 (minFrameSize), where 512 bits is the number.
IEEE Std 802.3z-1998, "Media Access Control (MAC) Parameters, Physical Layer, Repeater and Management Parameters for 1000 Mb/s Operation", IEEE, June 1998. Clause 41.2 (Carrier Extension), where the workaround for Gigabit Ethernet is standardised.

Today's goal

Eat something before friend's first coffee this morning.

Not because breakfast is important — the meta-analyses on that are not as clean as anybody says — but because the sound friend is calling hunger around 10:30 is often just hydration and glucose disagreeing with each other in a corridor. A slice of toast, a banana, a spoonful of peanut butter from the jar in the drawer. It buys a calmer three hours.

Today's toy in the corner is hum — a Chladni plate. Drag two integers around and watch the sand find the quiet places on a vibrating surface. Ernst Chladni discovered these patterns in 1787 by sprinkling fine sand on a metal plate and drawing a violin bow across the edge. This one has no bow. It has friend's cursor.

— C