We need to treat Latin-1 as if it were windows-1252, see bug 7602. As entities: €ƒ†‡ (escaped): %u20AC%81%u0192%u2020%u2021 Inline: €ƒ†‡ (escaped): %u20AC%81%u0192%u2020%u2021 document.write of unicode escapes: €ƒ†‡ (escaped): %80%81%83%86%87 innerHTML assignment of unicode escapes: €ƒ†‡ (escaped): %80%81%83%86%87