https://bugzilla.redhat.com/show_bug.cgi?id=1833858
--- Comment #23 from Luke Hutchison luke.hutch@gmail.com --- PS Droid Sans Fallback is often selected as the sans fallback font by Chrome on Fedora, and since the droid-sans-fonts package is default on Fedora (or at least a dependency of many other packages), this font frequently breaks Web browsing in Korean on Fedora -- so it's a pretty visible bug.
Even within a single sentence, some words (words starting with 가) fall back to Droid Sans Fallback, and some fall back to Noto Sans CJK. See a depiction of this pattern here:
https://github.com/harfbuzz/harfbuzz/issues/3003
The reason for this is probably that 가 is the pronunciation of the first jamo in the Korean alphabet, so it is used when spelling out bullet points etc. (equivalent to A, B, C, ...), and I'm guessing that Droid Sans Fallback contains the phonetic spelling of each jamo such as 가. So the fallback decision is made based on the first character in each word (at least in Google Docs, which probably calls the font measuring API for each word in a sentence, triggering a fallback decision for each word).
This indicates that the fallback logic can be improved. Currently the fallback determination seems to be being made based on the first Unicode codepoint in a span of text submitted for shaping. Is that a correct assumption?
i18n-bugs@lists.fedoraproject.org