How did they relicense without one?

They did not relicense.

All HashiCorp have done is made their own future contributions subject to BSL, not relicensed any previous contribution, nor do they have the ability to do that without copyright assignment of each and every contribution, which they do not have (specifically I have never signed a copyright assignment, and nor has at least one other of the top ten contributors of all time).

Nothing about MPLv2 prevents inclusion in a commercial product provided the attribution and file-based copyleft provisions are followed.

Is that something you can do? I wouldn't have thought there previous license would be compatible with the BSL, and so distributing the software under the BSL wouldn't be possible?

It's standard for BSD and MIT style licenses and similar to just require the copyright notice be retained in e.g. a LICENSE file or equivalent.

A bunch of commercial software right now (cough I'm writing this on MacOS) uses BSD-licensed foundations.

MPLv2 is file-based copyleft, and there are no restrictions on including it in a commercial work provided that changes to MPLv2-licensed files are accessible per the terms.

Indeed it’s something of a spiritual successor to CDDL which was designed precisely for that purpose (although reliable sources also say it was designed to be GPL-incompatible, so that is literally a he-said/she-said case).

The CDDL is an MPLv1 derivative, so similarities should be unsurprising

> All HashiCorp have done is made their own future contributions subject to BSL, not relicensed any previous contribution,

There's are only 2 .go files left in the hashicorp/terraform.repo which still have MPL-2.0 in their license headers, so either there was no community code left, so either there are virtually no such contributions left in Terraform or they have in fact attempted to relicense code.

This looks like relicensing to me: https://github.com/hashicorp/terraform/commit/53c34ff49cfbc1...

Btw, no file that you still show up in `git blame` for has an MPL notice header.

I will follow up on that, it was not the case even quite recently. If they really have rewritten all code, then great - I have no problem with this!

Replacing code whose license you don't want is definitely a fair-and-square way to manage a project and change the overall terms under which you can distribute it. (As it happens, that tactic is often essential to efforts to open-source formerly proprietary projects, too.)

*Provided that the replacement is not simply a rewritten version of the same code that achieves the same result through overwhelmingly similar methodology.

The function of code is explicitly not protected by copyright and the breadth of this exemption was clarified and expanded by Google vs Oracle.

The LibreOffice project faced a similar dilemma and have carefully kept the old licenses in their files, and new license in new code in new files.

what license did they change to/from?

How indeed.

Directly answered by [1].

[1]: https://news.ycombinator.com/item?id=37266016

Answered incorrectly. All the code is relicensed with BUSL header on each and every file.

Changing the license in the repo does not change the license of files you have already distributed under the MPL. That would require them to revoke all those licenses, which they can't do under the terms of that license that they've already agreed to. This entire thread and OpenTF, and maybe all of OSS itself would probably not exist if that were not the case.

In the official repo I can still check out an older commit with the MPL on it, which means they are still distributing code under that license, but it wouldn't matter if that were the case or not under the law.

edit: In fact, software as a business ( and many other professions ) would not exist if license terms could be changed or revoked after distribution. License changes for new software to be distributed can change, sure. But short of a breach of contract as outlined in the license itself you cannot change a license you have granted.

That's not what they mean. They went through every file in the repository and updated the license to BUSL from MPL.

Unless they have approval from each contributor for a given file (or the contributor signed a CTA or CLA), they cannot re-license that particular file.

It's still MPL and only once all contributors approve or the MPL code is removed can it be re-licensed to an incompatible license.

Of course if they had wanted to move to a compatible license, for a lot of FOSS licenses that is an option. Regardless of whether that is an option for the MPL, they are moving to an incompatible license and could not do this.

----

So their only options for files that they can't convert to the BUSL from the MPL are:

1. Keep contributions to those files under the MPL but require contributors to sign a CTA or CLA guaranteeing those contributions can later be relicensed.

2. Move BUSL contributions out of the file and only keep the MPL contributions in that file until they've all been replaced.

Doing otherwise is violating the MPL.

Again, wrong. Whether that has been done is independent of whether that is permissible, based on copyright assignment.

Are you implying they had no right to do that?

Strictly speaking it is on Hashicorp to prove that they had the right to re-license the code. Which would require listing all contributors and whether a given contributor gave written agreement to re-license the code or whether they had signed a CTA or CLA.

To my knowledge they have not done that so anybody whose code is being used without permission could fight them on copyright infringement.

And worth noting is that they don't have a particularly small number of contributors. Terraform has over 1700 contributors and many of the other repos have over 1000 contributors or close to 1000 contributors (unsure of the overlap).

If they only started requiring a CLA or CTA later into their development, they probably haven't gotten permission from all the rights holders.

Remove the code from the repo, put it in another repo, probably. I don’t remember signing a CLA 7 years ago, but somehow they relicensed without me needing to agree. All my code was pulled out into a separate package/plugin that is still MIT.