[dikei]

.NET is a managed-language with a VM. In such language, a memory error in managed-code will often trigger a jump back to the VM, where they can attempt to recover from there.

For native code, there's no such safety net. Likewise, even for managed language, an error in the interpreter code will still crash the VM, since there's nothing to fallback to anymore.

[wizofaus]

True, if you're talking unrestricted native code, I'd essentially agree with the OP's implication that only the OS (and the CPU itself) is capable of providing that sort of memory protection. I guess I was just wondering what something like AppDomains in C might even look like (e.g. all global variables are implicitly "thread_local"), and how much could be done at compile-time using tools to prevent potentially "dangerous" memory accesses. I've never looked at the postgres source in any detail so I'm likely underestimating the difficulty of it.