And we do have that as well -- it's everything attached to the work on the other side of "based on" or whatever other relationship is used. Like I said before, linear notes have a certain structure because they're inherantly flat and can only ever contain one level of data. We are inherantly hiearchical, so we shouldn't be bound into replicating the conventions required by flat data.
I'm also not entirely sure the legal worry is valid, though I do understand a bit better where it comes from. We plaster any question about commercial use with cautions that we don't guarantee data correctness, and there are several places on the site we have the same disclaimer. If anyone even wanted to sue us over that, I'm pretty sure our response would be "Sorry, we didn't know. Give us a minute to fix that for you." which is a better defense against the courts than conforming to any particular structure will ever be.
And finally, I get the reasoning of "there's no way to say which parts of the original were used in the derived work", but I also don't like the idea of duplicating the data. What if the upper-level credits are incomplete, or even wrong? Is someone adding the better data expected to check on every linked work in case they also need updating? If we figure we need that clarity, wouldn't it be better to add a couple relationship attributes to say "lyrics translated from" or "later version of music from" without involving the other half of the work?