> this lookup table of extensions to mimetypes doesn't feel like GPL "software".

Copyright nor the GPL are limited to software, collections of data are copyrightable as well; and thus they can fall under the GPL as well.

> To create a non-GPL version, you would have to do what? Research extensions without letting your eyes see this GPLed list?

Yes.

Data in and of itself cannot be copyrighted under U.S. copyright law: "creative arrangement" of it can be, but given that the data in question was generated from XML file, I don't think you could make a claim that the arrangement was copied.

? I don't follow - it seems pretty clear that this falls under the GPL from my reading of what is copyrightable, ie. they can't copy your compilation of the data.

Your reading of what is copyrightable may not be entirely in accord with the US Supreme Court's, which has ruled that mere compilations of factual data (a telephone directory in the case that set the precedent) are not copyrightable.

https://en.wikipedia.org/wiki/Feist_Publications,_Inc.,_v._R....

If someone is trying to apply the GPL to stuff which isn't legally copyrightable in the first place (as may be the case here), then their copyright isn't enforceable, and neither is the GPL.

There are definitely jurisdictions where this has been found not to be the case. In Australia the protection of databases is incredibly murky[1].

Telstra, the privatized government telecom company, failed to protect the White Pages data from another company that, if I remember correctly, had put it on CD (ah those were the days).

[1]: https://www.mondaq.com/australia/copyright/290668/can-a-data...

> collections of data are copyrightable as well

Collections of data are sometimes copyrightable. Depending on the jurisdiction, it may depend on the details of the collection.

I agree. Lists of facts are not eligible for copyright in the US, most famously phone books. It’s debatable as to whether or not this file is pure facts but I’m having a hard time seeing it as a “creative expression”.

Or use another source that is non-GPL - that's proposed here: https://github.com/rails/rails/issues/41750#issuecomment-805...

vs this https://gitlab.freedesktop.org/xdg/shared-mime-info/-/blob/m...

In a twist of irony, the software for which the copyright claim breaking rails was made is hosted on the free edition of gitlab, which is based on rails.

And according to the twitter-bio of the individual, who brought this up, he's related to Red Hat, which are also affected [^1].

[^1]https://github.com/RedHatInsights/compliance-backend/pull/79...