[btilly]

Again, that does not seem to have been the case here.

Again? Not sure where you said it. But the copyright holders in question are the authors of shared-mime-info, and they certainly never gave permission for their work to be used by Rails in the way that it was.

It depends where you are, which jurisdiction gets applied. Might explain the different expectation.

I'm in the USA. But I'm pretty sure that what I said is generically true.

It'd be the gem author that created an unlicensed derivative work, not anyone else directly.

Copyright is triggered by downloading unlicensed copies. And lots of people other than the gem author did that.

An unlicensed derivative was created by anyone who used Rails and wrote code that did mime detection - for example they were handling uploaded files.

It is an open question whether these cases are worth litigating, and what would be decided in court. They might well decide that there isn't enough creative work in the compilation for the file in question to have copyright protection at all. But in the meantime it would be a generally good idea to treat the issue seriously, and to accept that lots of people are potentially liable here. (Even if, in all probability, none will suffer more than a temporary inconvenience as the dependency is removed.)

[onli]

> Again? Not sure where you said it

Here, it was in the comment (and not an edit :) ):

> It would be further be complicated by the file in question being a database file. You typically can not license databases in a meaningful way under GPL. Even if you could, reading a GPL'd database has no chance of carrying GPL code obligations over to the consuming program.

But I actually just wrote again because I made that point in another subthread, no criticism implied.

[btilly]

Not so fast in that claim.

First of all the infringing file is https://github.com/minad/mimemagic/blob/master/script/freede.... Sure, it is in XML. But it contains a tremendous amount of free-form text, specific sets of pattern matching rules for the data types, and so on. It is a compilation of sometimes original research on the best ways to detect file types. Ruby has other mime libraries. The reason why this one was chosen is that its detection algorithms make better choices. And the reason that they make better choices is that they copied the decision rules from a GPLed project.

But even if it were a simple compilation, it still is not guaranteed that there is no copyright. See https://en.wikipedia.org/wiki/Copyright_in_compilation for an introductory article on what can and can't be copyrighted about a compilation. And one of the elements that matters is creativity in the selection of the material. A set of rules with a lot of "look for this" while leaving out various reasonable thats that don't work so well shows considerable creativity.

That said, a judge may decide otherwise. You never know until a judge decides. But I would not presume that there is no copyright interest to be had here.