> file semantics have really weird edge cases that you only find after losing data), so the new code is adjusted to actually be as comprehensive as the original

> file semantics have really weird edge cases that you only find after losing data), so the new code is adjusted to actually be as comprehensive as the original, at which point the new code is as ugly as the original I think the GP was getting at the fact that Rust/rustc handles many of these edge cases at the language/compiler/standard-library level, instead of at the application level. Which would mean, at least in theory, the new code would actually not have to be as ugly as the original in many cases.