This seems like a good application for a publicly auditable, immutable data store. Artist can birth a hash of the work into store along with their metadata. Post production steps would build on that record, creating a derivative work with new metadata. Labels, ditto. Distribution contracts go on top of that. When it finally gets a performance, that could be a transaction on top of that. Anyone can follow a performance back to its creators.
There are so many technical and non-technical problems with this, I don't even know where to start.
What's a hash of the work? Lyrics? Original score sheet (if there is any)? Any combination of the two? The first studio recording? Some songs may list a dozen collaborators, how's that attribution gonna work?
What's a hash? A hash of the resulting wav? mp3? flac? alac? any combination of the above? hashes of individual instrument tracks?
At which point does a song stop being derivative and becomes an original song (or music in general, see "Dies Irae" for example [1] or even Star Wars [2]). Who and how keeps track of attributions when lyrics are written by one person (or people), music by another, and performed by dozens of different performers over the years? How are collections tracked and attributed (an opera is a collection of musical performances viewed as a single work)?
These are just some of the problems people that try to track and attribute music face today. And they are not going to be magically solved by a "publicly auditable, immutable data store".
Nobody's claiming a magic bullet. Nothing you mention invalidates the idea, I think.
For the first part(the nature of the "hash"), there should obviously be standards to address these questions. An official recording could be identified by a hash of the actual recording in some standard format(e.g. raw wave files).
For distribution of score sheet and lyrics, those can be treated separately and have their own hash identifier according to their own standard.
The rest are philosophical and legal questions that are orthogonal to the technical problem.
The point of the suggestion, is to address the technical side of the problem.
Having an open platform where authors/copyright owners can hold public records of their copyrighted works and their distribution, and point to it as legal evidence, would certainly be useful, don't you think?
> The rest are philosophical and legal questions that are orthogonal to the technical problem. The point of the suggestion, is to address the technical side of the problem.
That's the problem: you're addressing the most boring and the least important part of the problem that has already been solved. No one in this world has any trouble tracking and attributing songs once a song's metadata is correct. And providing correct metadata is the entirety of the problem which you just glossed over as "philosophical and legal questions that are orthogonal to the technical problem". They are not.
Immutability is quite nearly the opposite of useful here. The most common way for data to be wrong is at the interface boundary: it was entered wrong in the first place.
Immutability, the way I think the author means it, doesn't mean data can't be changed ever, in any way. It just means that modifications(e.g. corrections) cannot completely overwrite previous values without leaving any trace, but instead have to appear as amendments, visible alongside the previous versions of the data.
If it was entered wrong, you can issue a correction, but a trace of the first, wrong value will stay available.
That means nobody can just say something was never what it was, just that it changed(and then have to justify why).
An interface should avoid unintentional mistakes. Easier to detect invalid signature or hash paired with other signature or hash versus an incorrect name or other ID.
A book is usually a single work attributable to a single author (or a group of authors). And you don't usually stream a book as "let's stream pages 175 to 185" [1]
In case of music assigning ISBNs to CDs doesn't really work because people more often than not listen to, stream and broadcast individual tracks. And there's no global ID system for individual tracks. This becomes even more messy when you consider that "Original performance 1983", "remastered 1994", "best of 2003" and "Japanese Christmas Special 2013" are all different tracks (even if it's the same track with the same song), often has different distributors and rights holders, and people will fight to the death defending a particular version :)
[1] Audiobooks are a slightly different beast, but they are still much easier to match to an author.
Yes, they’re called ISRC ids. But while they’re used to uniquely identify audio recordings, they don’t solve the problem of metadata being wrong or getting stripped during transfers
This is a great idea in theory, but how do you come up with an immutable hash of a song that a user can enter into the database? Even if you have a recording, most music these days is stored in a lossy format. If you have the MP3 version, the Apple AAC version, and some high quality .flac version, they're all going to sound mostly the same, but the underlying data is not even close to identical.
How do you hash the audible result, in a reproducable way? I'm not trying to bash on the idea mind, I just think this is an interesting problem and I'm curious if solutions already exist. Surely Shazam and the like have to be doing something in this vein.
