The standard is not vague on conformance, I agree. However, I think
there are parts of the standard that appear to contradict each other,
if not in words, then certainly in intent (speaking of which, perhaps
we should ask Charles?).
I think we agreed that browsers should all handle this the same way
(though I don't really even feel strongly about that, except in the
default case), but we most certainly did not agree on what that
means. I am afraid that by introducing language now, we will be
limiting our options in the future (for example, I think it desireable
to indicate the error in some other way, or to be able to display
glyphs even if the system doesn't directly support them).
Anyway, this is a client-side issue, and people have said time and
again that the HTML-WG should not be specifying client implementation
details, which is precisely what this will do:
if(character_value > 255) {
output_string(number_to_charref(character_value));
}