Doing font substitution on PUA code points was causing problems, because we have found a lot of fonts have garbage entries in their cmaps in the PUA, due to the implementation details of certain font-editing applications (which use the PUA part of the cmap as a "scratch" area).
Did this cause real problems for users?
Yes. There are an alarming number of fonts out there that have garbage entries in the PUA region of their cmap. Not to mention the GB 18030 and HK-SCS fonts that have PUA entries that *cannot* be removed until the relevant characters are encoded in Unicode (don't get me started on that decision...). We cannot decree what font vendors do: the fonts are out there, and some of the foundries that produced them aren't even in business any more. We cannot force end users to upgrade their entire font collection to eliminate less-than-squeaky-clean fonts (there would be a lot of empty Fonts folders). Users tend to consider such problems to be OS problems, not font problems, however unfair that may be.
The system was basically picking a random font in many cases. Rather than have a font substitution mechanism that picks a random font, we disabled font substitution for PUA characters.
But I suggest that that "randomness" can actually be controlled by users
- by font installation or code point re-assignment.
We always try to keep the needs of non-expert users foremost. We continue to refine our implementation of font substitution; we will keep this issue in mind and look for ways to accommodate expert users of the PUA.
Deborah Goldsmith Manager, Fonts / Unicode liaison Apple Computer, Inc. [EMAIL PROTECTED]
