[License-review] License Review Submission: Irrevocable MIT License (MIT-I)

[Rob Landley]

On 8/15/25 13:52, Jean-Sebastien Carle wrote:
> Thank you everyone for taking the time to review my proposal.
 > What I'm essentially trying to create is a MIT style license
 > where the same license cannot suddenly change its license to go
 > from being free open source software to paid open/closed source
 > software.

The permission grant on the existing code remains on the existing code, 
somebody else could host it (and fork it) under the old license.

Any license that DOESN'T allow you (or anyone else) to do a proprietary 
fork is a "viral" copyleft license like GPL. Plenty of those exist, and 
people have their reasons for not using them.

For example, a company I worked with prototyped a hardware project with 
GPLv2 fonts but can't ship that because their chip's ROM would then be 
in violation of GPLv2 unless they not only maintained a site hosting the 
source (which is STILL DUBIOUS because Stallman and Bradley and so are 
are such EPIC DICKS ala 
https://www.linux.com/news/gpl-requirement-could-have-chilling-effect-derivative-distros/ 
) and even if they did, what impact would it have on chip resellers in 
different countries (where does/doesn't first sale doctrine apply) and 
just generally they will not be going there. Linux in replaceable board 
firmware is funky enough, but GPL in hardware where you've signed NDAs 
about ASIC PDKs for mask creation and Bradley keeps suing people over 
toolchains used to create vendor firmware images? Just... no. Even if 
YOU'RE crazy, way too much money involved in making a mask for lawyers 
at both ends NOT to sign off on it, and they never would.

 > Trust in building software using open source software is being
 > decimated by a large quantity of highly used, high profile,
 > dependencies which developers build into their own software on the
 > premise that that dependency is in fact free open source software.

Dependencies suck. Have fewer of them. You're not going to fix that with 
licensing.

It doesn't matter what license python3 is under if python software 
regularly refuses to run on a 6 month old version of the interpreter:

https://landley.net/notes-2023.html#21-11-2023

https://web.archive.org/web/20241128031739/https://www.spinics.net/lists/linux-sh/msg64907.html

My toybox project's only external dependency is "libc". (Ok, and various 
linux /proc interfaces.) I did https://landley.net/aboriginal/about.html 
(and am doing https://github.com/landley/toybox/tree/0.8.12/mkroot) to 
produce a reproducible Linux environment that provides all its own 
dependencies. (It's a pity my "building the simplest possible linux 
system" talk was so jetlagged, I should redo it...)

http://lists.landley.net/pipermail/toybox-landley.net/2020-July/011898.html

 > Then, after these dependencies have become tightly coupled and
 > embedded in upstream software products, those dependencies have
 > switched to restrictive, paid, licenses which essentially hold the
 > consumers of those dependencies hostage.

Yes, it sucks. That kind of enshittification was basically what GPL was 
trying to address 30 years ago.

> I personally write all my software that I share as open source packages
 > under the MIT license to give all consumers of my packages the freedom

I use 0BSD because picking a public domain adjacent license instead of a 
public domain equivalent license leads to the stuttering problem.

https://landley.net/toybox/license.html

But you do you...

 > to use them as best as they see fit without any restriction
 > whatsoever. I would love to be able to license my packages under an
 > MIT like license such as the one I'm proposing to be able to say to
 > consumers: "This dependency is open source software, free, and without
 > restrictions. You may build your own software using these dependencies
 > with the peace of mind that the attached license cannot be revoked nor
 > substituted for a more restrictive one, now or in the future."

You can do that now. Even the original copyright holder cannot REVOKE 
previously granted permissions if the license didn't have a clause 
letting them do that (in the USA this is called promissory estoppel), 
they can only change the terms on NEW stuff.

They can take down their github, or check poison pill commits into it, 
and when they do you get https://asterisk.dynevor.org/xfree-forked.html 
and 
https://www.linux-magazine.com/Online/Features/LibreOffice-vs-OpenOffice 
and so on.

This is not a new issue.

Rob