You should probably take a look at URI::Find and consider mixing it with HTML::TokeParser (or friends). It will be more robust and you'll have fun, one hopes, exploring what those packages can do.