... sleepless night full of programming and debugging

I didn't sleep last night. There is a nasty bug in the Java XML parser. Under certain circumstances (inherited default namespace with together with attributes present) the parser presents a rich imagination. In this case an element namespace properties are wrong. No namespace URI is assigned to the element and the null namespace is attached.
As a result when serialized extra namespace definition appears xmlns="" and as a result the element is not bound to any namespace. A colleague of mine performed an excellent analysis of the problem.

Here is a result of the sleepless night (traverseDom method walk the tree recursively calling processNode on each node:

traverseDom(doc.getDocumentElement(), 
                new NodeProcessor() {
                  public void processNode(Node n) {
                  //TODO watch this in case the java xml parser will get better in future versions
                  //TODO watch this for general purpose XML signature processing where such elements may be legal
                  //fix parser error - we can afford this as no element is unqualified
                  // in case there is a unqualified element with some attributes with default null namespace
                  // it is concidered as an error produced by buggy parser
    if (n.getNodeType()==n.ELEMENT_NODE && 
                      n.getAttributes().getLength()!=0 && 
                      n.isDefaultNamespace(null) && 
                      n.getNamespaceURI()==null){
          String prfx=n.lookupPrefix(n.getParentNode().getNamespaceURI())
                        n.getOwnerDocument().renameNode(n,
                        n.getParentNode().getNamespaceURI(), prfx+":"+n.getLocalName());
                  }
                }
              }, true);