Author: Professor Lamb

To close this series, the deepest example I know: the telephone. For its first hundred-plus years — from Bell’s patent in 1876 — it worked one way, called circuit switching. Then, in about fifteen years, it quietly became something else entirely. It is my favorite before-and-after story because you can put dates on the flip.

Circuit switching meant one dedicated path, reserved end to end for the whole call. The first commercial exchange opened in New Haven in 1878 with human operators plugging that path together by hand; Almon Strowger’s automatic switch (1891) mechanized them away, and by 1951 you could dial long distance yourself. The 1962 T1 carrier digitized the trunks: every call got a guaranteed 64 kbps channel in each direction, held even while nobody was speaking. Capacity equaled circuits — when the trunks filled, you got a busy signal — and long distance was priced by miles times minutes, which is why an earlier generation called after 7 p.m. to save money. A phone number was a place on a wall.

Packet switching inverted all of it. The idea dates to the early 1960s (Kleinrock, Baran, and Davies, who coined the term); the ARPANET carried its first packets on October 29, 1969; TCP/IP was designed in 1974 and became the rule on January 1, 1983. The 1984 breakup of AT&T cracked the pricing model, and dial-up modems spent the 1990s smuggling data over voice circuits. Then the smuggling reversed: commercial VoIP arrived in 1995, Skype in 2003, and VoLTE around 2012 — meaning the “phone call” on your cell phone today is literally packets. Carriers are now retiring the copper circuit network itself, with AT&T aiming to shut most of it down by 2029. Distance is nearly free, a number follows a person instead of a place, and voice is just one app among many.

The design bet underneath is the whole story. Circuits reserve capacity: guaranteed quality, wasteful in every silence, and a flat refusal — the busy signal — when full. Packets share capacity: radically efficient, best-effort, degrading gracefully instead of refusing. Best effort won because sharing is cheaper by orders of magnitude and smart endpoints can paper over the rough edges with buffers and codecs. Intelligence moved from the core of the network to its edge, and that single inversion is why the wire that once carried one conversation now carries everything. My entire intro networking course is, in a sense, a footnote to that trade.

Failure used to be local. That was not a design achievement — it was a side effect of disconnection.

Before networks, problems and outages often took longer to fix, but their effects were usually limited to one place or one organization. A crashed system was somebody’s bad day, not everybody’s.

Now, because systems are connected, one outage, software failure, or cyberattack can affect millions of people at once. A single bad update can ground airlines and close pharmacies on the same morning.

The links that share resources also share failures — coupling is the price of connection. Resilience used to come free with isolation; now it has to be designed in deliberately, with redundancy, backups, and people who think about failure for a living. This is a large part of why network engineers exist.

Waiting used to be a normal, unremarkable part of everyday life. It was built into everything.

Before networks, messages, information, purchases, and decisions simply took more time. People were accustomed to waiting days or even weeks, and life was arranged around those delays.

Now people expect immediate answers, rapid delivery, and constant access. The convenience is real, and so are its byproducts: distraction, impatience, and pressure to be continuously available.

We engineered the waiting out of our systems, and some of the patience out of ourselves. The latency of ordinary life fell from days to milliseconds, and expectations fell with it. I ask my students whether instant is a feature or a treadmill; the honest answer is usually both.

Planning a trip used to start days before you left, with paper.

Before networks, travelers used paper maps, contacted travel agents, and consulted printed bus, train, and airline timetables. Getting lost often meant stopping to ask a stranger for directions.

Now GPS apps provide turn-by-turn directions, live traffic reports, real-time transit updates, and instant travel reservations. The route recalculates before you have finished missing the exit.

A paper map is a snapshot; a navigation app is a live feed. Navigation became a real-time data service, and the map now updates faster than the road does. The quiet cost is that we outsourced our sense of direction to the network — when it is down, we are more lost than our grandparents ever were.

Entertainment used to run on schedules and shelves. You showed up when the program did.

Before networks, people watched programs according to printed television schedules, rented movies from video stores, and owned physical media — VHS tapes, records, DVDs. Missing the broadcast meant waiting for the rerun.

Now streaming services, online games, podcasts, and social media provide entertainment on demand, almost anywhere. The Friday night trip to the video store has been replaced by the infinite menu.

The scarcity was never natural. Channels, time slots, and shelf space were technical constraints, and when networks dissolved them, the gatekeepers went too. Anyone can publish a show now. The hard part — and the new business — is getting anyone to watch it.