Okay, I interpreted something Eliot said to mean that given an
architectural DTD like this:
<!-- superclass elements -->
<!ELEMENT OUTER O O (INTRODUCED1)>
<!ELEMENT INTRODUCED1 O O (INNER)>
<!ELEMENT INNER O O (#PCDATA)>
The existence of OUTER and INNER architectural elements could force the
existence of an INTRODUCED1 architectural element as if those elements
were directly parsed. The idea seemed bizarre at the time so I should be
happy it isn't there. It seemed as if it would be useful for "wrapping"
one element in another. Architectural forms make more sense as an
application-level inheritance mechanism without it.
Paul Prescod