Or you could include these regional conventions in the character encoding, so that each file knows how it represent its numbers, and enforce them consistently throughout applications.

That would be a easy programmatic way to solve the problem, that wouldn't require the whole world to adopt a single standard. But of course it'll never happen, as it would require that all programmers know their shit and follow good practices, when it's much easier to throw the technical debt onto users and let them cope with the problem.