Summary

  • RFC 1926 defines a complete four-bit character map, a start character, Morse modulation and seven carrier frequencies, but it does not define how a receiver recognizes the end of one frame.
  • Contemporary minimal link specifications show that small does not mean implicit: even SLIP states its delimiter and escaping rules, while PPP framing states synchronization, transparency, integrity and invalid-frame behavior.
  • The historical lesson is not that acoustic IP was impossible. It is that a reversible sender recipe becomes an interoperable protocol only when receiver state and failure boundaries are shared and testable.

What follows the first beep?

Imagine the receiver has detected the Morse letter b, the designated start signal. More letters follow. Each belongs to a sixteen-character alphabet. Then there is silence.

Does the silence close the frame? Is it merely the gap between Morse elements or characters? Did interference erase the next tone? Did the sender stop halfway through an IP header? Should the receiver deliver what it has, wait, discard it, or keep appending the next transmission?

The puzzle is not imported from a modern protocol checklist. It is created by the asymmetry inside RFC 1926 itself. The two-page memo gives the sender a precise little procession. Divide the datagram into four-bit chunks in “network beep order”; map all sixteen possible values to letters; prepend b; key a steady tone in ordinary Morse code. Choose among seven “Acoustical Signature” frequencies, from 440 to 784 Hz, so several “Local Acoustical Networks” might coexist. For normal operation, use 440 Hz.

The reception section then compresses the other half of the link into one sentence: perform the process backwards.

Backward processing can invert the character table. It cannot, by itself, tell a receiver which observed sounds constitute one valid character stream. The map answers “which four bits does this letter represent?” only after framing and timing have answered “which letter was heard?” and “which letters belong to this datagram?”

A joke with an exact documentary status

The memo was dated 1 April 1996 and marked Informational. Its status notice says it specifies no Internet standard of any kind. The RFC Editor’s current catalogue places it in the Independent Stream. That current label should not be projected backwards as proof that every feature of the modern stream process already existed in identical form in 1996.

The genre is nevertheless well documented. RFC 8700’s history of the series describes April 1 RFCs as a special Independent Stream practice: humorous submissions without a formal technical review and approval process. The wordplay is structural—ATM becomes Acoustical Transmission Media, LAN becomes Local Acoustical Network, and an AS number becomes a musical pitch. The humor depends on looking just complete enough that the missing engineering becomes visible.

That status matters because an RFC number proves archival publication, not interoperability. The frozen sources do not report an independent implementation, a conformance exchange, a noisy-room test or an operating deployment for RFC 1926. They also do not prove that nobody ever experimented with it. The responsible conclusion is narrower: the document alone does not settle the common receiver choices.

Minimum specifications can be explicit

The comparison with small real protocols is revealing. RFC 1055’s description of SLIP calls the protocol a sequence of characters that frames IP packets “and nothing more”. It openly says that SLIP supplies no addressing, packet-type identification, error correction or compression. Yet it defines an END character, escapes END and ESC when they occur in data, recommends starting with END to flush noise, supplies sending and receiving logic, and gives a practical maximum datagram recommendation.

That is minimum specification as discipline. What is omitted is named; what must agree at both endpoints is stated.

RFC 1662’s HDLC-like framing for PPP makes the receiver boundary still clearer. A flag marks frame beginning or end. Escaping or bit stuffing prevents payload data from impersonating the flag. A Frame Check Sequence detects corruption. Invalid frames and inter-frame fill have defined treatment. These mechanisms are not mandatory design choices for every acoustic experiment; they are evidence that “reverse the sender” leaves several independent receiver decisions unresolved.

The title’s second joke also has a serious foil. Actual Classical IP and ARP over Asynchronous Transfer Mode defines an interoperable environment over AAL5: virtual-connection assumptions, default LLC/SNAP encapsulation, a 9180-octet IP MTU, address resolution, an end-of-PDU indication in the last cell, and an explicit statement that higher layers handle retransmission. Completeness does not mean that every layer solves every problem. It means the boundary and the responsible layer are visible.

Six steps from alphabet to operation

RFC 1926 is easiest to read as an evidence ladder.

  1. A representation maps each nibble to a letter.
  2. A modulation turns letters into keyed tone.
  3. A framing contract lets a receiver delimit one datagram.
  4. Integrity and recovery rules make truncation, corruption, loss and duplication observable.
  5. Interoperability evidence shows independently built endpoints exchanging normal and impaired test vectors.
  6. Operational evidence measures throughput, latency, loss, range, coexistence and failure behavior.

The memo supplies the first two rungs and one opening clue for the third. Its phrase “ordinary Morse code” does not state dot duration, tolerance, resynchronization or the treatment of competing senders. Its seven pitches label possible shared channels but do not define contention or collision recovery. Its crowded-area security advice is a warning, not a mechanism.

This is where Running-Code Primacy becomes a useful later test: publication is not operational reality; implementation, validation, deployment and use create new evidence. Minimum Initial Specification adds the crucial refinement that minimum does not mean vague. Rules needed for interoperability must be strict even when they are few. Reality Layers separates the durable symbol—the RFC, the joke, the alphabet—from the executable fact of a waveform being framed, checked and delivered.

These essays came later and do not describe Johnny Eriksson’s intent. They help us ask the right historical question. RFC 1926 did not fail a deployment test in the evidence available here; it stopped before defining the test.