He's kinda not wrong (that he's wrong) but he kinda is (because some of it is not entirely wrong)<p>A qualifier about pre-network communications is that the constraints of the copper wire telephone were writ large in how people did it: the copper was capable of doing mostly square wave signal but it was pretty horrid, it had been designed to do analogue voice, amplitude modification of volts down the line, 12v to 48v class, with a bizarre DC back voltage to power the local handset. Doing square wave was frowned on. It made other phone users angry and it risked breaking the equipment. Don't do it!<p>That constraint "don't stuff up the switching equipment and wires in the CAN" is why ADSL speeds and the whole "64 kilobit maximum on a subrate" and "1.44mbit on a T1/E1 link" (don't get me started on 56/64 and 1.44 vs 2) are defined by things as prosaic as the length of a copper wire run in metro uptown New York, close to where Bell technicians were allowed into the pits: if they could make the runs between manholes work properly, they could reliably guarantee that service for any run of copper in their network.<p>Prior to this, devices like card readers and hollerith punches worked by tweaking individual signals. The interface in some respect was "parallel" not "serial" because you had to hold a line 3.3v or 5v high (or low) to tell the computer you wanted to send data, and then toggle data on other lines. Clocking and things like that was enforced down all of the wires which were the 25 wire RS-232 standard. Thats ok when you have 25 wires. If you push it the other side of the phone network, you don't: you have basically 2: read and write. You have to invent things to send complex signals down a simpler path. Ok, there is a period of time where you beg TWO phone lines, and use them in parallel so you have two control signal lines and two datalines, but thats kind of not going to scale: the phone network was designed for as few wires per house as possible (the cable they used to use had far more pairs, because they expected to lose some and maybe sell a second phone line in the future)<p>Getting things to work asynchronously, auto-clocking down the data wire by using manchester encoding, working out how to do non-return-to-zero, all that stuff was smart people who had worked on Radio, Radar, Cryptography during the war at Bletchley and Bell Labs and the MIT Radiation labs.<p>We still leverage a lot of this underneath fiber systems. They still do trellis encoding, multiple bits per baud, huffman codes, manchester. Ethernet now, isn't what Boggs and Metcalfe wired it to be then. But we still call it ethernet.<p>There is also scaling effects. what can we do at scale, with the technology we paid for already because we really don't want to build it out again. The capex on the phone network was a 50 year bond model. Nobody wanted to overbuild it.<p>When I came into the industry punch cards were still in use. When I got into the telco space, there were regulatory limits to how many muxes you could use to compress voice into the same width of pipe (a concept called a QDU or quanticized unit of distortion was used to limit it to 2:1 compression)<p>Recent HN stories have looked at how the US national microwave backbone still exists, as a fallback, and as an alternate tower framework for entrepreneurs to play with in high speed datacomms. They built some of it for cold war defence reasons. Selling calls NY to SF was almost a secondary goal, given you had copper phones anyway.<p>A lot of this is buried under the "history"<p>(This is also all kinda wrong of course.)