MVP and EMiC will be needing to do XML diffing in order to produce critical apparatus/variant encoding, so I've started looking at the options. Juxta looks fanciest, and promises to handle TEI files, but I was unable to get it to accept a second file once I'd loaded one into a collation; that just seems to be a bug, but even so the output it creates is HTML and not very useful. I found several other programs (stored in ~/Applications/xml_diff_tools) and managed to get two of them working pretty well, one called diffxml and one called fuego (there's also one from IBM which is very old, and wouldn't run). The two that worked produce potentially useful XML output files which could be processed into something a human could use to check a list of differences and encode variants.