A spectre is haunting Unicode

The comments mostly landed on two extensions of that idea. First, several people pointed out that ghost characters are not a weird Unicode bug so much as a consequence of how CJK writing systems were standardized in the first place. Big historical source sets like the Kangxi Dictionary already contained dubious entries, and Unicode inherited a lot of that complexity from earlier efforts rather than inventing it fresh. Second, the conversation quickly widened into Han unification, because the same gap between “character,” “glyph,” and local usage still causes practical pain. The useful takeaway was not philosophical hand-wringing about purity. It was that East Asian text processing remains full of cases where visually similar forms differ by locale, semantically close forms have different code points, and identical code points can render in culturally wrong ways if the font stack or language tagging is off. A smaller but sharp subthread clarified one of the article’s examples. Commenters said the ghost character 彁 likely came from a bad digitization of 彊 in a Taisho-era newspaper, not from a real printed use of 彁 itself. That reinforced the core point of the article: once an error enters a respected source or index, later systems may preserve it as if it were authentic. People also noted that this is exactly why OCR, search, and cleanup workflows around CJK archives remain brittle. Even when the underlying text is “wrong,” the wrong form can become a stable identifier that downstream systems must handle.