Relicense under dual MIT/Apache-2.0

[marwes]

Relicense under dual MIT/Apache-2.0

[hawkw]

I license past and future contributions under the dual MIT/Apache-2.0 license, allowing licensees to chose either at their option.

[Marwes]

@cmr After reading through the Apache license I don't have any objections to the contents. I do however dislike that there needs to be a header in each file. There isn't some version of Apache which does not require this redundancy?

[emberian]

@Marwes I do not believe that the license header is required, only that it is preserved if it exists. Having a LICENSE-APACHE file (or similar) also fulfills the requirement "as indicated by a copyright notice that is included in or attached to the work".

[emberian]

e.g., rust-lang/rust#21481 (comment)

[Marwes]

Re-read the part at the end and that would be how I understand it as well now. In that case dual licensing is fine by me, marking as ok.

[Marwes]

@ildarsharafutdinov commented in the duplicate issue so I marked him here.

[skade]

I license past and future contributions under the dual MIT/Apache-2.0 license, allowing licensees to chose either at their option

(against my personal opinion outlined in skade/leveldb#18, but I don't want to stand in the maintainers decisions way)

[aochagavia]

I license past and future contributions under the dual MIT/Apache-2.0 license, allowing licensees to chose either at their option.

[Marwes]

I guess I should mention that I wouldn't mind staying as MIT so don't feel forced to accept this in any way. The only reason I feel tempted to do this license change is that I was under the impression that MIT did not require the license included in binary distributions (I would of course appreciate attribution but I don't want to force it). Thus changing the license to allow this would be more in line with my actual intent even if it complicates things (the Apache license is 8x the size of MIT which is annoying).

@skade I agree that the argument that this increases interoperability within the rust eco-system is a bit "suspect".

[skade]

@Marwes Well, you would still ask for a copy of the APL included alongside in a binary distribution. It's just that the MIT text includes the author name, so one copy per project must be included.

I also don't see how that's a problem, as legalities mean that distributors have to look at every project anyways on distribution. Appending another MIT blob to the resulting licensing document doesn't seem much more work then making sure that a copy of APL suffices to fulfill the requirements.

[Marwes]

@skade Then I misunderstood again and I am much less inclined to do this change. If a license need to be included anyway I don't see how it is any more complicated to just make sure the MIT license from each library used than to make sure that at least one Apache license is included.

(It occurs to me that a cargo sub command which traverses all dependencies and generates an attribution file would make MIT and Apache an equal amount of work anyway).

[emberian]

(Note: APL in the SPDX database is the Adaptive Public License. The Apache license is sometimes refered to as the ASL (Apache Software License), but this name is obsolete)

@Marwes One must already include an Apache license from using the Rust stdlib. It's true that a cargo subcommand would make this easier. It's up to you!

[skade]

@cmr Still, the checking whether the current license note fulfills all distribution requirements must occur, independent of the result (either that the Apache 2.0 license already included suffices or that another MIT blob must be included).

And that's where the work is.

[emberian]

Yes, definitely.

[Marwes]

I don't see enough value coming from this change so I will close this. Some good discussion came out of this though which at least made me learn a thing or two about the intricacies of both MIT and the Apache license.