Fun read. My take on the "special something" that causes people to care about implementation is that you want to specifically control parts of a computer. Yes

Fun read. My take on the "special something" that causes people to care about implementation is that you want to specifically control parts of a computer. Yes, this is typically for run-time performance; but crucially it is because you want to specifically reason about what the computer will be doing to implement your program. And the more that your computer is capable of doing, the more you are likely to care about different aspects of how it executes. I'd hazard this goes both ways, too. That is, it is not just in terms of how much the computer is capable of doing, but what the real world systems being modeled can do. For a fun example, you can make a sorting network that can sort integers on a computer rapidly, but that doesn't necessarily help you sort a physical quantity of anything. Is akin to making a computer that can win a game of chess, but only if the human moves the pieces for it. All too often, people present programs as if they are truths independent of either model or computer. Which is doubly wrong as they are constrained by the model of the language they are implemented in, as well as being dependent on the model of whatever problem they are describing.