> difference. This *ought* to work! One simple solution, which I just
> tested, is to make <pause> etc. a member of model.global.
I think I will add model.divPart.spoken to macro.paraContent for now.
Otherwise, it would be nice to make the content of the <seg> dependent
on whether or not it has an ancestor::u, but that is as far as I can see
not possible in plain relaxng. Maybe, schematron could be useful for this?