None of the above. It is just more convenient to either use a library (that is already nice, shiny and debugged) or some already preprocessed data (which the PDML mykl talked about probably is. Also might be safer to parse because of added redundancy. And there are lots of XML parsers ready to use).

Parsing binary data is not really hard, but the program tends to be a bit more unreadable and cryptic than a text parser because you have to use constants like '\x02'