This is a follow-up to Jon’s original post on Carefully (but purposefully) oxidising Ubuntu and Julian’s migration spec for 25.10. We promised transparency throughout this process, and this post is written in that spirit. What happened after the announcement Following the decision to adopt rust-coreutils, we got to work. Any package shipped by default in Ubuntu must be promoted to Ubuntu Main, which requires passing a thorough security review. We quickly assembled an internal team spanning Ubun...
I don’t think I’ll ever understand the constant complaints about the license. If it were the kernel or some software that was particularly unique, then I’d understand. However, there are many existing implementations of the coreutils programs that are already under permissive licenses. If someone didn’t want to use the GPL, they could just use one of those. This is partly why it is incredibly fiddly to write cross-platform shell scripts.
The mit license allows someone (some company) to modify the open source codebase and sell the result without making their modifications public.
It allows the software equivalent of the enclosure of the commons.
If there was a particularly large or significant and widespread codebase —like for example the coreutils— that was used everywhere and mit licensed, a company could make their own slightly different coreutils without publicizing the differences and use their position in the market to enclose the commons of knowledge about the use of that software. Such a situation would lead to a fractured feature ecosystem and confusion around best practices. In that environment, the biggest and most popular software distributor would benefit because their product would be most common and therefore the best target to design around.
I know there’s a lot of “coulds” and “woulds” in that sentence, but that’s exactly what happened in the 80s and 90s with the ostensibly open source Unix codebase and the reason why the gpl was invented.
It’s already fractured, as I literally mentioned. That’s why it’s hard to write cross-platform scripts. Part of the reason it’s fractured is that the implementations most commonly in use other than GNU coreutils are permissively licensed and thus cannot easily adopt unique features from GNU coreutils.
In any case, at this point, changing the coreutils license itself will not materially change much in terms of how fractured the existing landscape is given that people could already use Busybox, Toybox, programs from any of the BSD userlands, etc. if they didn’t want to use GNU coreutils for whatever reason.
If it doesn’t matter then why not use the original projects license?
I know you’re not able to read minds or responsible for the greater rust community but how come when I or anyone else asks the above question of any mit licensed rust project is the answer never “huh, I guess if the license doesn’t matter then we can gpl it no problem!” And always “no, and get your politics out of my code!”
It clearly matters to someone because everyone’s feet are always dug in to the sand about sticking with mit.
Is rust-coreutils being developed by Canonical? Then it sounds like shooting themselves in the foot. Why give competitors a chance to take over a vital package that is at the core of their OS?
mit lets companies take them without contributing back critical stuff like security fixes.
their money and resources are very important to keep foss alive and this relies a lot on the gpl because it just means they are forced to take some responsibility for the projects they use to make their billions.
That’s great, except they could already just use a permissively licensed implementation. This is in fact what a lot of companies already do. For instance, Android uses Toybox, macOS uses utilities originally ripped from NetBSD (mostly), etc.
Generally, a lot of companies also don’t contribute back fixes upstream. They’ll often just dump the code in some hidden away corner of their site as a giant source blob.
For something like coreutils, where a significant change is sort of unlikely in the first place, thinking the GPL makes a difference is bizarre to me.
Because it was started as a project to learn Rust by one dude.
Also, that was back when Rust had bad documentation (at least a couple years before 1.0), so by far the easiest way to learn was by making something like this and looking through other existing projects like the compiler or Servo.
That doesn’t answer the question why use different license than the original. And why not change the license/fork to gpl when it became more than a fun project. As we see it is a major issue with the project.
Being able to take someone else’s code used as a learning exercise for your own learning without worrying about it being GPL’d is quite useful. You seem to be arguing permissive licenses should never be used, which I think is ridiculous. A project meant to just learn about XYZ language/framework/whatever by implementing “simple” tasks is one of the most basic examples of a project that should be under a permissive license.
The only thing that could realistically be done is to license all changes going forward as GPL. If someone wanted to fork the project to do something like that, they could. But of course no one will bother, because the people who are terminally rabid online about this project not being under the GPL contribute to neither this project nor GNU coreutils.
It is not a major issue. It’s only really an “issue” at all because people who don’t contribute and likely would never contribute anyway constantly complain about it. I will state this again: there are multiple already existing implementations of the coreutils programs, so there is practically nothing keeping companies tied to it. There is little actual benefit to the coreutils programs in particular being under the GPL.
I don’t think I’ll ever understand the constant complaints about the license. If it were the kernel or some software that was particularly unique, then I’d understand. However, there are many existing implementations of the coreutils programs that are already under permissive licenses. If someone didn’t want to use the GPL, they could just use one of those. This is partly why it is incredibly fiddly to write cross-platform shell scripts.
The mit license allows someone (some company) to modify the open source codebase and sell the result without making their modifications public.
It allows the software equivalent of the enclosure of the commons.
If there was a particularly large or significant and widespread codebase —like for example the coreutils— that was used everywhere and mit licensed, a company could make their own slightly different coreutils without publicizing the differences and use their position in the market to enclose the commons of knowledge about the use of that software. Such a situation would lead to a fractured feature ecosystem and confusion around best practices. In that environment, the biggest and most popular software distributor would benefit because their product would be most common and therefore the best target to design around.
I know there’s a lot of “coulds” and “woulds” in that sentence, but that’s exactly what happened in the 80s and 90s with the ostensibly open source Unix codebase and the reason why the gpl was invented.
It’s already fractured, as I literally mentioned. That’s why it’s hard to write cross-platform scripts. Part of the reason it’s fractured is that the implementations most commonly in use other than GNU coreutils are permissively licensed and thus cannot easily adopt unique features from GNU coreutils.
In any case, at this point, changing the coreutils license itself will not materially change much in terms of how fractured the existing landscape is given that people could already use Busybox, Toybox, programs from any of the BSD userlands, etc. if they didn’t want to use GNU coreutils for whatever reason.
If it doesn’t matter then why not use the original projects license?
I know you’re not able to read minds or responsible for the greater rust community but how come when I or anyone else asks the above question of any mit licensed rust project is the answer never “huh, I guess if the license doesn’t matter then we can gpl it no problem!” And always “no, and get your politics out of my code!”
It clearly matters to someone because everyone’s feet are always dug in to the sand about sticking with mit.
Is rust-coreutils being developed by Canonical? Then it sounds like shooting themselves in the foot. Why give competitors a chance to take over a vital package that is at the core of their OS?
Some Canonical employees are working on it but it’s not originally a Canonical project.
mit lets companies take them without contributing back critical stuff like security fixes.
their money and resources are very important to keep foss alive and this relies a lot on the gpl because it just means they are forced to take some responsibility for the projects they use to make their billions.
That’s great, except they could already just use a permissively licensed implementation. This is in fact what a lot of companies already do. For instance, Android uses Toybox, macOS uses utilities originally ripped from NetBSD (mostly), etc.
Generally, a lot of companies also don’t contribute back fixes upstream. They’ll often just dump the code in some hidden away corner of their site as a giant source blob.
For something like coreutils, where a significant change is sort of unlikely in the first place, thinking the GPL makes a difference is bizarre to me.
So why did they choose to use permissive license instead of the license of the original?
Because it was started as a project to learn Rust by one dude.
Also, that was back when Rust had bad documentation (at least a couple years before 1.0), so by far the easiest way to learn was by making something like this and looking through other existing projects like the compiler or Servo.
That doesn’t answer the question why use different license than the original. And why not change the license/fork to gpl when it became more than a fun project. As we see it is a major issue with the project.
Being able to take someone else’s code used as a learning exercise for your own learning without worrying about it being GPL’d is quite useful. You seem to be arguing permissive licenses should never be used, which I think is ridiculous. A project meant to just learn about XYZ language/framework/whatever by implementing “simple” tasks is one of the most basic examples of a project that should be under a permissive license.
The only thing that could realistically be done is to license all changes going forward as GPL. If someone wanted to fork the project to do something like that, they could. But of course no one will bother, because the people who are terminally rabid online about this project not being under the GPL contribute to neither this project nor GNU coreutils.
It is not a major issue. It’s only really an “issue” at all because people who don’t contribute and likely would never contribute anyway constantly complain about it. I will state this again: there are multiple already existing implementations of the coreutils programs, so there is practically nothing keeping companies tied to it. There is little actual benefit to the coreutils programs in particular being under the GPL.