It's not exactly mirrored even. The way the 24 break down is this: You need 2 power and 2 ground pins per side to carry the large currents needed. Those all ge

It's not exactly mirrored even. The way the 24 break down is this: You need 2 power and 2 ground pins per side to carry the large currents needed. Those all get wired to the same (thick) cable (though usually split in two to make the cable more flexible). That's 8/24 pins, 2 wires total. You have two pins (per side) to signal which mode the cable is being used for and which way around it's plugged in. Because there are so many alternate modes with each having its own signaling standard, you can't pack those in-band, particularly for the case of analog audio, which doesn't have a digital data frame at all. Two are connected across from one connector to the other, so there's only three new wires for the 4 pins. So that's 12/24 pins, 5 wires total. You have two pins per side for USB 2.0. Those get wired to one side only, but are connected together in the device (not the cable). The opposite side pins are unconnected. So two connected and two unconnected new pins per side, and two new wires. 16/24 pins, 7 wires total. Some cables have active transceivers in them, and to ensure that the far-side transceiver is powered, there's an extra wire carrying its supply. This brings us to 8 wires. The remaining 8 pins are mapped to four high-speed differential pairs. The important thing about these is that they ONLY support high-speed low-voltage differential data and are protocol-agnostic. To get the enormous bandwidth the standard needs they have to split them up into four different serial lanes. So it's not a single serial bus. It's 5 serial buses, a handful of power and configuration pins. This brings us up to the full 24 pins (22 of which are in use) and 16 wires (or 18 if you opt for the thinner, doubled-up power and ground wires). It's very much not a serial bus. As for optical cables, that would work, but then you need to transfer all the control signals in-band and either stick to a single signaling standard (and have to do high-speed conversion for everything else) or somehow enclose various protocols in a single frame (like ethernet). This would require a fundamental redesign of peripherals, rather than a small change (a mux and control IC and new port). Additionally, optical transceivers are very expensive and power-hungry at high speeds, to the point where even for extremely high link speeds copper is still being used for ethernet. TOSLINK succeeded because at the very low data rates it has, the optical link hardware is cheap and doesn't use much power.