Grammar caching with multiple schemas for the same target namespace

Hi,

I'm recursively processing DITA maps using the SAXParser with 
XMLGrammarPool. All files are in the default namespace but have 
different XML schemas associated (set my @xsi:noNamespaceSchemaLocation).

The problem is that the XMLGrammarPool as well as the SAXParser assumes 
that the schema only depends on the targetNamespace. Thus, they will use 
the same schema for a topic file as it used for a map file.

I already implemented my own XMLGrammarPool class that checks if two 
schema descriptions are equal based on their system id. But it seems 
that the parser stores a mapping between namespace and systemId. Thus, 
it asks the grammar pool for a schema with the map system id when 
processing a topic.

My current workaround is to create a new instance of the SAXParser for 
each file I'm processing. This works fine, but I'm wondering if I could 
improve the performance by reusing the same instance for all files. But 
I couldn't find out how to change the behavior for getting the grammar!?

Thanks for any hints.

Regards,
Patrik