why is entity ref expanded in the internal subset (and related questions)?...

I'm having a problem with the following...

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE vxml SYSTEM "voicexml1-0.dtd" [
	<!ENTITY % BigEntity SYSTEM "BigEntity.ent">
	%BigEntity;
]>
<vxml version="1.0">
	<form id="init">
		<block>
			&BigEntity;
		</block>
	</form>
</vxml>

BigEntity.ent looks something like this...

<?xml version="1.0" encoding="UTF-8"?>
<!ENTITY BigEntity '
		<prompt>another prompt</prompt>
		<prompt>another prompt</prompt>
		<prompt>another prompt</prompt>
		<prompt>another prompt</prompt>
		<prompt>another prompt</prompt>
		<prompt>another prompt</prompt>
		<prompt>another prompt</prompt>
'>

When internalEntityDecl(String name, XMLString text, XMLString 
nonNormalizedText, Augmentations augs) gets called, both the 
text.toString() and nonNormalizedText.toString() contain the fully 
expanded contents of the entity (a bunch of <prompt>another 
prompt</prompt> entries).  After I build this into a DOM and 
serialize it, it looks like...

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE vxml SYSTEM "voicexml1-0.dtd"
  [
<!ENTITY %BigEntity SYSTEM "BigEntity.ent">
<!ENTITY BigEntity "
		<prompt>another prompt</prompt>
		<prompt>another prompt</prompt>
		<prompt>another prompt</prompt>
		<prompt>another prompt</prompt>
		<prompt>another prompt</prompt>
		<prompt>another prompt</prompt>
		<prompt>another prompt</prompt>

">
]>
<vxml version="1.0"><form id="init"><block>&BigEntity;</block></form></vxml>

Why are the contents expanded?  Why not just give me "%BigEntity;" 
from the original document?  Is there any recourse?

A second problem is that reparsing this file gives me the following 
parsing error...

test9.gen.xml:4: Error: White space is required after "<!ENTITY" in 
the entity declaration.

The parse error seems to be with the external entity declaration 
"%BigEntity".  If I change that to "% BigEntity", the parse error 
goes away.  But, the "name" parameter passed to me in the 
externalEntityDecl() method is exactly "%BigEntity".  If the parser 
knows it can't deal with that, why doesn't it pass me "% 
BigEntity"?  Seems odd that I'd be forced to split the "%" apart from 
the rest of the entiy name manually.


thanks,

Jake