>Are there any known cases of processing that *depends* on this being
>a WF error (and would thus be surprised if it someday became allowed)?
Well, XSLT and dozens of other applications depend on entities having
been expanded, so in effect they rely on undefined entities being a
well-formedness error (ie the kind of error that stops normal
processing). At the very least, their behaviour is implementation
dependent if there are undefined entities, whether it's a
well-formedness or validity error.
I'm not sure what you're getting at here. If undefined entities were
merely a validity error, how would experimentation be more tractable?
And why isn't adding a dummy doctype good enough?
In theory, something like
<!DOCTYPE foo [<!ENTITY % x ""> %x;]>
is sufficient to make undefined entities a validity error, though
a quick check suggests that not all parsers believe that.
-- Richard
