So a user agent is "allowed" (because it is outside the scope of the
standard), for purposes of display, to treat any string, anywhere, that
begins with =? and ends with ?= as an RFC 2047 token, and decode it.
seems like a stretch. and 2047 takes exception to that.
while it's generally the case that IETF standards do not specify UIs,
I don't recall any rule that limits IETF standards' applicability in this way.
people keep citing it (without references) as if it were one.
in general I think it's wise for IETF to avoid specifying UIs, but occasionally
it's also wise to make an exception to that.
Keith