- added the patch from Carl Nygard <cnygard@bellatlantic.net>
   which allow impressive speed improvement on dataset with large text
   pieces, but at the cost of broken binary compatibility and slightly
   bigger memory usage.  Configure with --with-buffers to activate them,
   they are protected with XML_USE_BUFFER_CONTENT define.
 - added xmlCleanupPredefinedEntities(), memory allocation cleanup
Daniel
11 files changed