Xerces XMLSerializer And Character Entities

Question

After struggling for hours trying to get the "portable" java transformation classes to output my DOM object to a file, I switched to Apache's XMLSerializer. The main problem I'd been having with the transform is that character entity references were maintained in the DOM tree, but kept getting resolved in the file (someone finally pointed me to a document stating that there is currently no way around this problem). The first time I tried using Apache's classes to write the DOM tree, it worked almost perfectly. The entity references actually showed up in the file; the only problem was a minor bug that caused notations to print wrong and make an invalid XML file(<!NOTATION tif SYSTEM 'tif'> became <!NOTATION SYSTEM 'tif'>). I downloaded Xerces 2.4.0 and the notation problem was fixed; unfortunately the character references are now being resolved AGAIN. I'm at the end of my rope....how can I prevent the XMLSerializer from converting references like &rarr; to bizarre utf-8 characters? Any help would be greatly appreciated.
Here's my code...
// xmlDocument - DOM Document object
// fileName - The name of the XML file to write to
OutputFormat format = new OutputFormat();
format.setLineSeparator( LineSeparator.Unix );
format.setIndenting( true );
format.setPreserveSpace( false );
XMLSerializer serializer = new XMLSerializer( new FileWriter( fileName ), format );
DOMSerializer ds = serializer.asDOMSerializer();
ds.serialize( xmlDocument );
P.S. - This exact code kept the references in the outputted XML file until I upgraded my xercesImpl.jar. I've also played with most of the methods on OutputFormat without any luck.

Answer 1

Here are a few snippets:
<!ENTITY rarr "#x30"> <!-- In our process this should NEVER be resolved -->
<value>Here is a character entity: &rarr;</value>
The rarr entity reference becoms an EntityReference object in the DOM tree, but after I run it through XMLSerializer the XML file looks like this:
<value>Here is a character entity: ?</value>