I am not yet convinced that <milestone> is an appropriate encoding
for this sort of thing. A <milestone> (IMHO, and I recognize there
are others) is not a generic point in the text. Rather, it divides
the text into two different measurements of the same thing. That is,
the stuff before the <milestone> is the same as the stuff after (at
least as far as that which the <milestone> is measuring), only at a
different navigational point. If it can be argued that <div> is
appropriate, then what occurs before & after is substantially
different: they are at different conceptual levels, as it were.
The theory "if the author intended this to be a true division it
would have been made that way with separate divisions with headers
and references from the Table of Contents" does not seem correct to
me. <head> is optional for a reason -- authors don't always use it;
printers make tables of contents that list only down to a certain
level of division every day.
So while I would not go so far as to claim <div> is *right*, it
certainly seems better than <milestone>. If it were really important
to avoid the effort of using <div> or <floatingText> for some reason
(like the end is hard to find, perhaps?), I would consider using
<space>, perhaps with some indication on rend= if there were
> But adding a new suggested value wouldn't affect the schemas (see
> above).... so I'm happy to consider suggestions for such values
> (complete with gloss), if they arrive in the next few days.
Actually, new suggested values do affect the schema. But just as with
re-arranging the order of <attDef>s in an <attList>, they do not
change the set of documents that are valid against the schema, and
thus might still slip in between the shutter slats.