Building Yomu: Safari, furigana, and why “plain text” often isn’t plain

This is the third part of a short series about building Yomu, an iOS app for reading Japanese text with adaptive furigana.
In the previous post, I described how the app tokenizes text, renders furigana, and works with an offline dictionary. This post focuses on a much lower-level problem: getting usable input text in the first place.
It turned out to be one of the most frustrating parts of the project.