Grouping tracks into works

Unfortunately, uniquely identifying compositions is hard. Is a composition a particular score edition, printing, original performance, recording? Even in fairly recent classical, it’s not so clear, just check hour many Stravinsky Petrushka’s there are. Are you sure that the one in that older CD you got is any of them, or some other variant? And when we get into jazz, with improvisation, all bets are off. The challenge of creating a reasonably complete and deduplicated composition database is not just technical but also critical: is this variant really distinct from that one? by what criteria? Much bigger and richer outfits than Roon have put a lot of resources into this problem with results that are still unsatisfying.

As another example, who are the performers in a particular recording? Amazing how many mistakes one finds, even with different releases of the same original recording.