UTR #50 revision 5 wrote:
There is actually one character for which a contextual determination would be useful and reliable: U+00AE ® REGISTERED SIGN, which can occur both following terms in kanji/kana and following terms in Latin. An occurrence of ® should be assigned the same class as the character it follows. Others? Enough to warrant the complexity of contextual rules?
I'm wondering if we had a value of "I" (inherited) if that would not work properly not only for the ® sign, but also any combining marks. You could then set NBSP as SVO=R and the Ideographic space as SVO=U to allow for exhibiting combining marks in isolation, and it would simply follow from their inherited value. We could even go with In/Ip (inherit next/previous) to accomodate open/close quotation marks. Does that seem like a reasonable solution to anyone else?