[ALT] Circuit-Switched Networking (AT&T Paradigm)

Circuit-switched networking — the telephone's model of establishing a dedicated end-to-end physical connection for the duration of a call — was the dominant telecommunications paradigm for a century before packet switching made the internet possible. AT&T's engineers actively argued that packet switching was technically inferior and commercially unviable.

The Fork

What won: Packet switching — breaking messages into discrete packets, each independently routed through a shared network, reassembled at the destination. Proposed independently by Paul Baran (RAND Corporation, 1962) for nuclear-survivable communication and Donald Davies (NPL, 1965) for efficient data networking. Implemented in ARPANET (1969), standardised as TCP/IP (Cerf & Kahn, 1974), and scaled to the internet.

What was abandoned: The circuit-switched paradigm for data networking — AT&T's Bell System engineers told ARPA researchers in the early 1960s that packet switching was technically inferior to circuit switching and would never scale. AT&T declined to build ARPANET. Larry Roberts (ARPA) was told by AT&T that his concepts were "technically impossible."

The Technical Argument AT&T Lost

Circuit switching dedicates bandwidth for the connection duration — efficient for voice (continuous, real-time), but massively wasteful for data (bursty, non-real-time). A circuit switched data network uses 1–2% of the allocated bandwidth on average. Packet switching shares capacity dynamically: 1,000 users each using 0.1% of a channel can share it without any one user noticing the others. The internet's efficiency over circuit switching is ~100×.

What AT&T Got Right (and Why They Resisted)

AT&T had a USD 50B+ infrastructure investment in circuit switching. Bell Labs engineers were not foolish — they correctly identified real challenges: packet delay variability (jitter), error recovery, routing complexity. All were eventually solved. But institutional self-interest, the NIH ("not invented here") syndrome, and regulatory capture (AT&T was a legally protected monopoly) meant they fought packet switching until they had no choice.

The Irony

Voice-over-IP (VoIP) — packets carrying voice calls — now carries virtually all telephone traffic worldwide. AT&T itself transmits its telephone calls as packets. The paradigm AT&T said could never work for voice became the only paradigm.

Current Status

Circuit switching survives only in niche applications: dedicated leased lines for ultra-low-latency financial trading (microseconds matter), 5G network slicing creates virtual circuit-switched channels over a packet network for real-time applications. The hybrid is packet switching infrastructure with circuit-switching emulation for quality-of-service guarantees.

Discovery Character

Surprise level: High — the internet's scale and economic impact were not predicted by anyone, including ARPA. Roberts himself expected a network of a few hundred computers.

Mode of the fork: Systematic technical argument (Baran's 11-volume RAND report was exhaustive) vs. institutional inertia. A rare case where the technically superior solution won despite the incumbent's opposition.