Or just collect mp3s? I'm guessing that at least happened while prototyping. And it seems like the most efficient way to make the service work in that era so it wouldn't surprise me if that data got used in production initially while handling licensing concerns separately.