i don't know of any way to get back the authentic html [perhaps a deep dump of the obj might reveal a secret cache], but - in theory -
you should be able to reconstruct the meat of it by 'serializing' the Document obj. i'll explain.
as i understand it,
java is able to do the rendering you are describing via model and view classes. the view is what gets rendered, the model is the nodal representation of the underlying html parse tree.
to make a long story short, i would try traversing "the structural portion of the subject that this view is mapped to" (ie. the document)
From the document root node (Element), you can traverse all the elements and attributes and 'serialize it' back to html.
there are classes to help with the serialization somewhere in javax.xml or org.xml .. you would use the relevant html dtd (or something simpler) for the schema.
i hope someone else can suggest a shorter way to do it...
hth
