As I see it, the details about presentation would only be
suitable for use in the DTD itself, i.e. they refer to the
a property/effect of a tag, not of the text. Thus, what I'm
suggesting is that a DTD could include supplemental information
for each tag which notes whether it has presentation effects.
Presumably such effects would operate as modifications to
a standing `vector' of such properties (as Scribe uses to
change states when it encounters a request to, say, indent
while inside a previous request to reset the left margin).
 
What seems possible is that a set of device independent
presentation capabilities could be determined. I`d prefer they
be numerically unspecified, i.e. an `indent' isn't a number
but a presentation effect described in terms of other
presentation traits such as a `left margin'.