I suspect that a lot of the inconvenience of modelling non-linear problems in Rust actually doesn’t come down to ownership semantics alone, but the combination of ownership semantics and the low-level memory model. However, I don’t have any solid evidence to support this, because ownership semantics have only really been implemented with the low-level memory model, because the low-level memory model strongly benefits from it (memory safety; it more closely matches how computers work internally at that low level) and GC languages don’t need ownership semantics, and people don’t generally restrict things without cause (and ownership semantics are definitely a restriction).

My thesis here is that Rc<RefCell<_>> and similar are an effective escape hatch to ownership semantics, but they’re much less convenient than a garbage-collected type in a scripting language specifically because of the language’s low-level memory model. A GC language with ownership semantics might be able to obviate RefCell, which is the cause of Rc<RefCell<_>>’s inconvenience.

I’d be interested to see experimentation with ownership semantics in languages with higher-level memory models (which broadly means GC languages). My thesis could turn out to be quite wrong, and I haven’t worked out all the details of how the combination would work (especially I’m hazy on parts of removing RefCell, which is pretty critical), but I think it could be worth some research—ownership semantics for its own sake, rather than ownership semantics for memory safety.

> 10. Getting used bypassing the borrow checker to reduce fighting implies people have just found a way to cope with the constraints it imposes.

Here’s the thing, though: it’s not just about finding a way to cope with the constraints it imposes; it’s also about thriving within the constraints it imposes, because it stops you from doing problematic things and guides you in the path of better designs much of the time (though non-linearity can definitely be a problem; it’s not all buttercups and daisies). When I work in JavaScript, I spend half the time missing ownership semantics, occasionally feeling downright miserable because ownership semantics would have made a problem far easier or less dangerous. Remember what I said about ownership semantics for its own sake rather than ownership semantics for memory safety? Yeah, most of the time I love Rust’s ownership semantics for the modelling rather than the memory safety.