As we are Germans there can be ü,ä,ö in COLLADA documents. The current revision of the libxml plug-in doesn't set the encoding attribute in the XML header. I changed that, but then the file was corrupted because the strings are interpreted wrong by the function xmlTextWriterWriteString. To fix that I added the conversion isolat1ToUTF8 if the preprocessor variable WIN32 is defined.
I don't know if this is the right way to handle this, but for Windows it works.