If you're happy to leave the XML documents in files, but want a system to help you locate specific documents, you might want to look at the SWISH-E indexing package. It comes with libxml integration so that you have full control over the indexing. For example you might choose to have the indexing process extract the contents of a particular tag and associate it with a named property in the index. Then, you can define a search to locate all documents with a specified value in that property. You can query the index from the command line and there is also a Perl module API.