Layering in Help Systems

Greetings Colleagues, and Happy 2009! 

I'm interested in everybody's thoughts on whether layering features in Help
systems can be applied globally based on semantic mark-up, or whether there
is a need for "ad hoc" layering. By layering, I mean features such as
popups, expansion links and dropdown links. By "ad hoc", I mean an author
choosing that this lump of text would be best presented to the user in a
dropdown, for example.

As an example, let's say we render a footnote <fn> in the HTML output from a
DITA collection as a popup. Can we apply a rule that says "in all cases (for
this Help system), <fn> elements will be rendered as popups"? Or will there
be an instance where we need to say "oh, THIS <fn> needs to be treated
differently". If we further assume that a transformation process somehow
links any <term> elements with a corresponding glossary entry into an
expansion link to that glossary entry, can we safely way that this will
happen in all instances within the help system?

I'm trying to explore whether we actually need different semantic elements
for "Help" features. It would be philosophical sound if we could avoid
having help-centric DITA elements that ended up embedding presentation
information in the content. Or having some type of DITA mark-up that only
applied to Help systems.

What do you all think?

Regards

Tony Self