The DOM tree in EML parser probably cause memory issue
This bug may be considered as an EML module bug.
Currently, EML parser (it will check validation about reference id in EML2) is
implemented by DOM tree. So if EML2 document is huge (has lots of inline data),
the DOM tree will be huge too and probably out of memory and swap.
So we need to reimplement the parser by SAX parser. This will avoid this issue.
MetaCat reuses the EML parser code, so this is a issue to Metacat too.