The engineering feats are indeed remarkable! But so were the engineering feats of the desktop PC era. Previously it was unbelievable that you could make a computer small enough to sit on somebody's desk. So those guys were learning how to miniaturize things too -- and somehow, along the way, they ended up building systems that had interchangeable parts.

The current crop of device manufacturers, on the other handed, punted on this goal. Is it really because this generation's engineers aren't good enough to solve an engineering problem? Is it coincidental that failure to solve this engineering problem results in higher profits?

I don't think that's the case -- I think the business planners in the current generation of manufacturers saw interchangeable components as a feature which would results in consumers saving money and buying fewer new phones, gave us the middle finger and never bothered to do it.

The history of computers is one of higher and higher levels of integration. Single transistors -> gate ICs on wiring boards -> processor with support chipset on PCB with peripherals on daughterboards -> system-on-a-chip. Each time it's improved performance and reduced cost.

> failure to solve this engineering problem results in higher profits?

Ultimately all engineering in a capitalist business is value engineering. Cost is not something that can be removed from the metric in people's heads which they use to choose solutions.

There are only two ways I can see out of this: a move towards lease-like arrangements for phones, where the manufacturer has an incentive for something other than obsolecence; or the EU pushing WEEE harder.

Well, maybe less of the engineering should be done by capitalist businesses. The more open platforms and protocols that the PC and Internet were built on usually weren't developed by private companies with profit motives. They came from researchers, hobbyists, foundations, and companies with other profit centers which weren't focused on squeezing every possible dollar out of that particular technology.

How we can encourage more of that is a tricky question. I certainly don't want a world where profit-motivated private companies can't build hardware and software. But I don't like the direction this current crop of businesses is taking us. I do think essential infrastructure doesn't belong in the hands of a single party -- we should have learned that lesson and internalized the risks of single vendor lock-in long ago.

You already had such computers, but the C64, Spectrum, BBC, Atari, Amiga and many others were following the same architectures we are going back to. It was the PC that lead the way into user customization, but the ever decreasing thin margins seem to have killed it

Except, as the parent mentioned, a lot less of those millions of transistors are still under your (the user's) control, compared to the PC era. The technological advances are indeed breathtaking, the amount of participation you have in those technological advances (except in a small number carefully pre-approved ways) is less so.

Also, I'm not sure we can make the "simply not economical" statement this early whitout having gotten a more detailed reasoning of the descision. After all, already many other comments in this thread point out ways in which modules could be useful that don't clash with architecture requirements. (e.g. exchangeable case). Also, as far as I know, Project ARA had a non-google "open source" predecessor. The people initiating that obviously found the idea realistic enough to explore.

Finally, if it were just technical constraits, we should see a golden era of modularisation in smart cars, as both available space and power are comparable to PCs. Yet, we see the same kinds of lockdown - or even more extreme forms - over there.