[tei-council] numbered divs

[Syd Bauman]

> does anyone feel that the looseness of this is unacceptable?

Yes, completely.


On the other hand, you voiced in earlier mail (not to the list) the
thought of further constraining this back to "only numbered OR only
un-numbered" with a Schematron rule. This may not be so terrible,
depending on what our end position is on the use of Schematron.


The general problem is that dropping all members of a class may make
some content models ambiguous in DTD (and maybe W3C) land.

I think it is important to seriously investigate the dummy-element
insertion idea, which could solve this general problem, not just this
instance. This is the idea that whenever the odd2dtd processor comes
across an empty class, it populates that class with one element
<dummy-do-not-use>. In DTDs this element wouldn't even be declared
(and thus, regular DTD validation should flag an occurrence of it as
an error in an instance). In any case (RelaxNG, DTD, W3C Schema), a
very simple additional Schematron rule would flag all
<dummy-do-not-use> elements as errors (following is untested
off-the-top-of-my-head):

  <s:pattern name="No dummy elements permitted">
    <s:rule context="tei:dummy-do-not-use">
      <s:report test="tei:dummy-do-not-use">
        The &lt;dummy-do-not-use> element should never be used. For the technical
        explanation of why this element exists see XXXX.
      </s:report>
    </s:rule>
  </s:pattern>