Ghost characters in Unicode are a window into how standards accumulate debt
The 'A spectre is haunting Unicode' piece got warm reception, partly because the author Paul McCann has credibility in Japanese NLP, and partly because the subject is genuinely strange: Unicode contains characters that were never real, copied from typographical errors in historical dictionaries. The Kangxi dictionary, which sourced most CJK characters, apparently contained significant numbers of 'ghost' characters, and Unicode inherited them.
The pattern: standards bodies, once they commit to backward compatibility, inherit every mistake made upstream. This is a clean example of how technical debt compounds across institutions, not just codebases. Unicode cannot remove these characters without breaking compatibility, so they persist.
This is a niche thread but it resonated because it illustrates a broader truth about infrastructure: the cost of fixing a mistake is often paid by people who had nothing to do with making it.
So what?
For founders building internationalization features, text processing tools, or anything that handles CJK character sets at scale, ghost characters are an edge case that can cause real bugs. More broadly, this is a reminder that standards are not clean abstractions: they carry historical baggage that will surface at the worst possible time.