This is a great talk by Sean Parent about defining and understanding what is generally meant by “Memory Safety” and safety research in general. Towards the middle, he shifts to talking about code correctness as the focus.
This talk starts off by referring to the NSA recommendations that were made regarding memory safety. It also covers other recommendations that have been made about shifting the burden of responsibility onto software developers. He responds to both of them.
This talk, even though presented at C++ Now, spends very little time on C++ and more on general issues surrounding safety. As such, there’s a lot here for all languages and programmers.
I was convinced long before I saw this talk that Turing Completeness implies that safety is an illusion. He also brought up an interesting point about how things like “don’t touch uninitialized memory” first requires a program to use uninitialized memory (that’s how it becomes initialized to begin with).
Overall, great talk.