On Mon, 23 Mar 2026 at 20:08, Marin Ivanov <metala@???> wrote:
> > Hmm, doesn't that just move the problem to a different level of
> > abstraction? Suppose, for example, that you made a perfectly
> > memory-safe implementation of the Ook language (which is Turing
> > complete), and then wrote a fully standards-compliant C compiler that
> > outputs Ook code. Does that mean that all C programs are now
> > memory-safe?
>
> I don't argue that "all C programs are memory-safe". My whole point is
> that memory-safety is not a property of a language, but of the
> program. In this case the C-Ook-object-code compiler produces
> memory-safe programs.

Memory-safe programs which (aside from abysmal performance overhead)
do the exact same thing as the original programs, since they were
compiled from those sources by a fully compliant C compiler.

So I suspect that either "memory-safe" can be true at one level and
not at another, or all C programs must have been memory-safe to start
with (which can't be true or this discussion wouldn't be happening).

ChrisA