Hacker News Reader

Comment by CivBase | Hacker News Reader

GitHub is still hosting the full youtube-dl version history, including versions which include those supposedly infringing tests. Does copyright law end with HEAD on master? Those tests are still there.

This makes it especially obvious that the RIAA's problem with youtube-dl was never really the tests.

View on HN

I've always wondered about copyright and HEAD.

For example, you can find a LOT of copyrighted font files that were committed somewhere in GitHub, and then removed in a later commit once they realized they'd accidentally uploaded a copyrighted file.

But they're still always there in the history, effortless to download.

I'm not really sure what to make of that. I don't think it would really count as removal in court... but it seems rare and complex enough that it's not worth bringing up?

If GitHub received a DMCA takedown notice, they would be obligated to takedown the copies of the fonts listed in the notice, including if old copies were listed. I'm unsure if they could say "all releases before X" or would need to link each one.

If the copyright owners tried to sue the project for copyright infringement, IANAL but I would assume that the removal from head would show an attempt to correct the mistake and limit liability.

If the copyright holder sued an individual I imagine it would matter if they were mirroring the repo or just intentionally downloaded the copyrighted files for personal use.

On that note, does git as a protocol even have a clean mechanism for redacting history like this? If someone were to press this to the logical extreme, how could a developer most cleanly excise violating history from a repo using current tooling?

There is a way, its not very clean, git filter-branch and you will have to force push all branches, which is fun with large teams.

Unfortunately in larger repos with long histories its extremely slow, and uses a lot of IO. I used it previously to clean up large binaries that were included early on in a repo's history, making it take up way more space than needed.

When you need to do it in a big repo, BFG Repo-Cleaner is usually the way to go: https://rtyley.github.io/bfg-repo-cleaner/

I assume you would have to revert to the parent of the offending commit, cherry-pick the non-offending code, commit, then rebase the entire master branch on that new commit.

Then you'd have to repeat the process for all forks and branches. It'd be a huge pain, but I think it's doable.

I've never tried something like this, though, so there might be some complications.

A lot of stuff like this is technically infringement but will never be worth anyone's time suing over, except maliciously.

> Does copyright law end with HEAD on master?

With regards to copyright law and "distribution", there's no distinction. The tests are still being "distributed", just from a different URL. If youtube-dl was in violation before, they still are now.

This is a confusing result. I would not expect any copyright litigant to sacrifice legal advantage for the sake of an adversary's convenience in maintaining complete version control history.

Could there possibly have been a miscommunication over what "remove the tests" meant? Or an offer of compromise outside of legal necessity? Or a bad-faith fulfillment of a promise to "remove the tests"?

Does the fact that the RIAA is packed with stodgy, old corporate lawyers who most likely lack even superficial understanding of the term "version control" lack appropriate explanatory power?

Such "stodgy, old corporate lawyers" are going to become extremely antagonized should they discover that somebody pulled a fast one on them. All you'd have to do is open a web browser and click around to show them that the tests were still accessible.

So what's confusing is the youtube-dl side's strategy. Are they really trying to pull a fast one? That would be incredibly unwise, so I doubt it.

Everything just points to the dmca takedown request being treated as if it was invalid; the repository was restored more quickly than needed and without any real changes.

I suppose it depends on how the claim is formulated? Take the CraftBukkit[1] claim for example. In that case, all commits since a specific point, plus all forks, were taken down:

"Pages including infringing content: [...] infringing as of commit [...] and every subsequent commit, including all forks that contain this commit [...] and all forks that share a common first commit [...] and every subsequent commit, including all forks that contain this commit"

This made any effort for restoration futile, since most of the repo was being claimed.

[1] https://github.com/github/dmca/blob/master/2014/2014-09-05-C...

> Does copyright law end with HEAD on master?

I think the RIAAs technical know-how ends with the github web interface, so in a way, yes, copyright effectively ends with HEAD.

Then when someone realizes to use the web interface to check out that old commit, the whole thing opens up again. Hopefully github's lawyers who OK'd this know a bit about git.

> This makes it especially obvious that the RIAA's problem with youtube-dl was never really the tests.

Or that the RIAA lawyers are technically oblivious

Or that they simply don't care about the specific infringement, rather they are just trying to chill the free-speech rights of software developers and fight a culture war on all fronts to expand the rights, privileges, power, and wealth of their clients.

Yes I really doubt the law cares about which commit some code is available on, as long as it is available.

> This makes it especially obvious that the RIAA's problem with youtube-dl was never really the tests.

It doesn't, because the RAII were not involved in youtube-dl's restoration at all.

I don't think the story ends here.

If the tests are the issue, they can just send another DMCA specific to those pages.

Either way, I think in the US anyone who has the most time and money wins, so… good luck with SLAPP.

Reply on news.ycombinator.com