I'm partial to XML::Rules. Others like XML::Twig. Either is good if the docs are large and you should be processing the doc as you read it. XML::LibXML can be appropriate also (XPath support is great if you need that sort of thing).