I must be confused by the entry for
<xsl:text/> (on page 323 it says the default value of d-o-e is "no"; on the
next page it says it's "yes").

My point is if a node-set contains only HTML markup, you can avoid the whole
output-escaping issue by setting output to HTML and using <xsl:copy-of/>.

Now I'm seriously confused. What do you mean by "a node-set
contains only HTML markup"? AFAIK a node set contains nodes,
and it doesn't make a difference which vocabulary the element
nodes are from. If you mean someone has a CDATA or something
with text which resembles HTML markup, then no, xsl:copy-of
wont convert this automagically into a node tree or whatever,
is stays a text node after being copied.