Incidentally, in my last job, I had to design a DTD with rather similar
requirements for fallback text.  We needed descriptions of the relevant
objects, without displaying the description where the object wasn't visible.

What we ended up doing was something like this, which won't be very handy
for XHTML, most likely... but who knows.

<alternatives>
    <alternative type="video/mpeg">
        <p>Here is a video of ...</p>
        <p><object type="video/mpeg" src="..."/></p>
    </alternative>

    <alternative type="image/png">
        <p>Here is a picture of ...</p>
        <p><object type="image/png" src="..."/></p>
    </alternative>

    <alternative>
        <p> I <em>would</em> have something to show you, but unfortunately
            it seems that I will have to speak a thousand words to write
            instead.</p>
    </alternative>
</alternative>

That enabled us to process which one to render without needing to look ahead
at all.  But in the case of XHTML2, there might be multiple objects inside
each section, and who knows how that would be handled (multiple types in the
'type' attribute on <alternative>?)

The current XHTML doesn't do enough to implement this, for the aforementioned
reasons (the description of the picture would still be rendered even if the
picture isn't there.)

TX

-- 
             Email: Trejkaz Xaoza <trejkaz@xaoza.net>
          Web site: http://xaoza.net/trejkaz/ 
         Jabber ID: trejkaz@jabber.xaoza.net
   GPG Fingerprint: 9EEB 97D7 8F7B 7977 F39F  A62C B8C7 BC8B 037E EA73