I dunno, people have been saying this about electronics in cars for longer than I've been alive, and I'm not that young.

I mean, first? finding a weird part is finding a weird part... and you are totally right that the electronics are the new weird parts and that way too much of the system is in the head unit, but... there have always been weird parts... and if you don't have a machinist (or these days an EE) friend helping you out, if you have a car with a weird part, you spend time in junkyards. (or the modern equivalents. Ebay has made finding weird parts so much easier)

My impression is that this is all as it has always been, except that it is easier to find weird parts than it was, and that if you want to make/rig a part, these days, you call up your EE friend rather than your machinist friend (and I have far more EE friends than machinist friends)

Me personally, I'm more comfy swapping out broken electronics modules than I am swapping out bolted-on metal parts, but either way, it seems like a pretty similar sort of thing.

I fail to see how an EE would be able to fix the problem with a trashed ecu. Once it's gone, it's gone, and there is no way of replacing it with a similar element without the producer know-how. The SW in those things is huge and you have great interconnectedness between the different systems.

I also couldn't machine a part for your carburetor.

My understanding is that the ECU controls the fuel injector, though, and that most of the inputs and outputs are fairly well known; and I'm told that it's a pretty simple control loop with the oxygen sensor, the TPS and mass airflow sensor. All in all, if it came down to it, I think I'd have better luck building an ECU from scratch than building a carburetor from scratch, after having fixed carburetors and fuel injected systems. Machining to those tolerances is hard.

(To be clear, I wouldn't put money on me personally doing either one of those things. I fix computers for EEs, and I'm no machinist. I'm just saying, micro-controllers aren't that hard if you are trying to do a simple thing... machining to carburetor tolerances really is.)

My understanding is that most cars don't have any sort of encryption between the ECU and the sensors; it's all stuff that someone with basic ee training could figure out with 20 year old equipment. If you have a popular/fancy car, people will sell you aftermarket/performance ECUs that are not recommended for use by the manufacturer, so while it's possible that it's a lot harder to do than I think it is, there's existence proof for the possibility.

I work in automotive embedded. And yes, the field is bloated. Yes, things could be more efficient. But even if all those problems went away, I still don't think it's feasible building one-off ECU's for usage in a friend's car.

Those things take a great amount of effort and time. The complexity is pretty big. Granted, I mostly work on instrument clusters, and I believe they are one of the biggest systems in the car, but the complexity is huge. I found it hard to carry the mental model of the whole system in my head.

After-market ECU's are deff possible, because you do a couple thousands of units at a time. But doing it for one ECU, I think not. Not if you put an upper-cap on the effort you are willing to expend. You're probably better off just buying a new car at that point.

Ya, I think early-mid 2000s are the sweet spot for cars:

The OBD port will tell you everything that went wrong, no SatNav and knobs/buttons everywhere.