encoding, charset and SOAP

I have a charset question.

It looks like some processors append the value of the submission's 
@encoding as the value of charset in the content type of the submission. 
  But in the 1.1 spec when it talks about SOAP, it says that if the 
mediatype contains a charset MIME parameter, it is appeneded.  So it 
sounds like there could be two charsets by the time it gets to the 
server.  Which is supposed to win?  Does SOAP mean that no encoding 
happens?  Or does the value of a charset MIME parameter mean THAT is how 
the serialization is encoded?

Any guidance welcome.

Thanks,
--Aaron