Conversation

The industry's attitude on memory safety is truly strange.

In the early 2000s, enforcing global hardening policies like stack canary, strict W^X, or ASLR was an extreme niche and highly unpopular work, because people think it's invasive and breaks compatibility. It remains so today, "Control Flow Integrity with Forward and Backward Edges", anyone? Then in the 2020s, rewriting entire C codebases in Rust suddenly becomes popular and mainstream in record time - although doing this breaks compatibility in its fullest sense - in contrast to the 15 years of delay on normalizing compiler hardening.

I guess it's because the latter group doesn't need to care much about compatibility at all, unlike the first group who must work with people who do care. So breaking compatibility is paradoxically more popular than preserving partial compatibility.