At one point or another you are going to need to know the structure of the XML you are going to work with anyway. The fact that DOM or some other maze-of-objects-based XML parser let's you get away with assuming a single occurance of some tag even though it may be repeated, while XML::Simple would cause your script to bomb out, is not always a good thing. I'd rather if my stuff failed noisily than if it'd produce incorrect results.

You can of course split the XML processing into two, completely unrelated phases. First, that knows nothing whatsoever about what the structure of the XML is and what data are you after and just blindly parses (well, it's not much more then lexing actually this way) the file into some kind of data/object structure. And second that does know where are the data you wanted and has to navigate the structure to get them. In my humble opinion this is often unnecessarily hard and inefficient.