That sounds significantly more like C++ trying to be a dialect of Rust, rather than the other way around. I don't think that was the GGP's main gripe.
But more importantly, Safe C++ is just not a thing yet. People seem to discount the herculean effort that was required to properly implement the borrow checker, the thousands of little problems that needed to be solved for it to be sound, not to mention a few really, really hard problems, like variance, lifetimes in higher-kinded trait bounds, generic associated types, and how lifetimes interact with a Hindley-Milner type system in general.
Not trying to discount Safe C++'s efforts of course. I really hope they, too, succeed. I also hope they manage to find a syntax that's less... what it is now.
I don't think Safe C++ has a Hindley-Milner type system? I think it's just the "Just the machine integers wearing funny hats†" types from C which were passed on to C++
In K&R C this very spartan type system makes some sense, there's no resources, you're on a tiny Unix machine, you'd otherwise be grateful for an assembler. In C++ it does look kinda silly, like an SUV with a lawnmower engine. Or one of those very complicated looking board games which turns out to just be Snakes and Ladders with more steps.
But I don't think Safe C++ fixes that anyhow.
† Technically maybe the C pointer types are not just the integers wearing a funny hat. That's one of many unresolved soundness bugs in the language, hence ISO/IEC DTS 6010 (which will some day become a TR)
No, Safe C++ does not have that type system. I was just trying to emphasize the amount of, let's be honest, downright genius that had to go into that lifetime specification and borrow checker implementation.
For C++, it'll be about cramming lifetimes into diamond-inheritance OOP, which... feels even harder.
Safe C sounds like a much, much more believable project, if such a proposal were to exist.
But more importantly, Safe C++ is just not a thing yet. People seem to discount the herculean effort that was required to properly implement the borrow checker, the thousands of little problems that needed to be solved for it to be sound, not to mention a few really, really hard problems, like variance, lifetimes in higher-kinded trait bounds, generic associated types, and how lifetimes interact with a Hindley-Milner type system in general.
Not trying to discount Safe C++'s efforts of course. I really hope they, too, succeed. I also hope they manage to find a syntax that's less... what it is now.