No, that’s not the proposal. The problem that Google has is that if it ships every possible Unicode character in a font, then that font will weigh many megabyte

No, that’s not the proposal. The problem that Google has is that if it ships every possible Unicode character in a font, then that font will weigh many megabytes and the vast majority of that will not be used. So they do “font chunking”: they split the font into smaller files that contain a subset of the characters, and serve it to users. If you look at Google’s stylesheet for Roboto you can see that the font is split into 7 chunks: https://fonts.googleapis.com/css2?family=Roboto&display=swap https://fonts.googleapis.com/css2?family=Roboto&display=swap Each chunk has a unicode-range hint associated with it, which tells the browser that it doesn’t need to download the font unless those Unicode characters are found in the markup it’s processing. The article’s assertion was that ₹ is a a common enough character among English speakers / writers that Google should move it into the base Latin chunk instead of putting it into the extended Latin chunk, which would save a bunch of font downloads. This is all a balancing act though. Every character you move into the “default” chunk makes that chunk (and the unicode-range definition) larger, so you have to be careful and work with data. And one imagines that Google has some pretty good data in this area.