Open litherum opened 2 years ago
Edit: Yes, I raised this in a previous issue: https://github.com/w3c/csswg-drafts/issues/2110
Agree, we do not do system fallback for PUA anymore in Chrome, it's bitten us before mainly when a particular icon or specific math font gets installed on the system and suddenly provides wide PUA coverage. On top of the correctness problems, searching for fallback in the PUA region is inefficient from a performance point of view, as it's so unlikely to find coverage in any font.
IIRC MS has a slightly different view on it, as they provide a mechanism called EUDC that places user-group wide extensions into the PUA range for specific system fonts: https://docs.microsoft.com/en-us/windows/win32/intl/eudc
See also @SergeyMalkin's earlier comment: https://github.com/w3c/csswg-drafts/issues/2110#issuecomment-352853932
CC @dlibby-
Interesting, thanks for the links.
Maybe we should say "certain installed fonts can be used for certain PUA characters, but UAs should try to be conservative" or something. This attempts to strike a balance between allowing for @SergeyMalkin's concerns while also attempting to not make the problem get any bigger in the future.
The spec currently says:
This should probably be made more strict. Because PUA characters are font-specific, and websites don't know what fonts are installed on the system (or any details about them like their character coverage), no installed font should match any PUA characters.
@svgeesus