Re tune comparisons :- another possible approach is to use abc2midi, mftext (supplied with abc2midi) and a scripting language, to extract the bare pitch&duration from a (ABC) tune and then turn that into something that can be compared using fuzzy text matching (Levenshtein). It doesn't work spectacularly well, but nor does anything else. The quality of the input ABC is a major drawback, as someone already pointed out. Variant repeat structures are a particularly misleading PITA.

Pamela, thank you for a fascinating thread.