Probably someone from the unleashed project can give a better answer, but my understanding is mostly unhappiness about the illumos contribution process.
The illumos RTI (request to integrate) process is designed to ensure a certain quality of contributions.
The downside of this is that it sometimes takes a long time to contribute a change.
For illumos there is also pretty high expectation of backwards compatibility.
What unleashed now offers is a place to do more radical changes and quickly iterate on new ideas.
While the small illumos community is already pretty fragmented my personal hope is that in the future work done by both projects keeps getting merged from time to time.
Maybe this also serves as a wakeup call to the illumos project to make some improvements to the contribution process itself.
These are not goals. Every OS in the world has contributions and releases. Linux has the goal to be pragmatic, BSDs have the goal to be technically perfect in a certain area, Windows had the goal to be on every desktop computer in the world and more, and OSX has the goal to be smooth and sharp.
If your highest goal is that someone contributes code to your OS why would anybody use it? Think about htat question and you might find your actual goal. If you can't find a goal, then maybe the existing OSes are sufficient. Choose one, live with it, and develop something else on top that doesn't have a satisfiable alternative yet.
If you've ever tried to contribute something to the Linux kernel you'd know it's not an easy process. They say that they want to "encourage new contributors to repeatedly contribute". So in this respect it is very different from the Linux kernel.
Now I will admit this is going to be tricky. How do you dissuade junk contributions or misfeatures but at the same time encourage contributions? They may end up with an overall declining quality which kills off other goals such as user adoption.
> If you've ever tried to contribute something to the Linux kernel you'd know it's not an easy process. They say that they want to "encourage new contributors to repeatedly contribute". So in this respect it is very different from the Linux kernel.
I have contributed a small 1-time change to the Linux kernel. It was relatively easy and straight forward; I sent my first patch to the wrong list, but the maintainers helpfully corrected me.
The difficult part is finding what to contribute, IMO.
I think you're forgetting the large learning curve to get to that point. You learned how to build the kernel, git, the kernel style guide, signed-off-bys, checkpatch, get-maintainer and finally running git-email. And there's the assumption that you knew what to change and how to fix it, but most kernel error messages are opaque - it can often be impossible to find out why a system call returns -EINVAL or where an error message comes from.
And that's just for a small change. Large and/or controversial changes often involve multiple rounds of review and can take (literally) years to get upstream. I'm sometimes amazed when I read LWN about how long it has taken for features to get integrated into the kernel.
I don't know what Unleashed OS is planning but it seems obvious we could make this a whole lot simpler.
> I think you're forgetting the large learning curve to get to that point.
Openness in general requires a large learning curve, and I think it's a mistake for any open source project to be marketing itself as a democratic entity that literally anyone can contribute to. If anything, open source is more like a technocratic government (ergo the "Benevolent Dictator For Life" position in a project). Even Wikipedia is a huge pain in the ass to add to at this point.
I've noticed a similar phenomenon with the open science movement, where scientists publicly release their data. Unsurprisingly, few folks from the general public usually work with open data with the necessary assumptions to draw valid conclusions from it. Other scientists, as insiders who've built up decades worth of expertise in statistics and other analytical techniques (at least in theory; in practice not always) are usually the only people who have the ability to fully milk the data for what it's worth. Science, however, is possibly even slower and more political than linux kernel development, although I don't have enough knowledge to make a proper comparison.
> You learned how to build the kernel, git, the kernel style guide, signed-off-bys, checkpatch, get-maintainer and finally running git-email.
These are all pretty well documented tasks. It might not be 'one button simple' like a gitlab merge request, but I thought it was pretty easy overall. It did take some time investment on my part to learn the submission steps (about a weekend total, including the compiling and testing locally).
> Large and/or controversial changes ...
I suspect if they are large or controversial, that in itself implies the barrier to merging is going to be higher. How many new contributors (non-paid contributors especially) want to merge something controversial on the first go?
Could it be simpler? Possibly. Sometimes complex systems do require complex processes for maintenance. I'm just saying I didn't think it was especially burdensome to contribute a small change as a brand-new, completely-unassisted by another person (before sending patch) contributor.
Everybody tries this. In the beginning contributions are easy. After almost 30 years of everybody developing on the same code base nothing is easy any more. That is normal and can't be avoided.
I agree though, that it might be good marketing to claim that. Especially younger people haven't seen that 3-5x already so they still think it might be possible. Just like a 4 Hour Workweek, or that you could be Fluent In 3 Months in a new language.
With no snarkiness intended, why would I want to use this? Is it just for people who want an OS to contribute to for fun? The only features really mentioned are "it's not horrifically legacy" and "it's easy to contribute to".
29 comments
[ 5.3 ms ] story [ 90.4 ms ] threadA public GitLab instance where everyone could submit pull requests would be really cool (or just host it on GitHub).
http://repo.or.cz/unleashed.git/blob/HEAD:/docs/illumos-refu...
I am not sure if I would use it, but I like this attitude!
Can someone explain why it was forked ?
The illumos RTI (request to integrate) process is designed to ensure a certain quality of contributions. The downside of this is that it sometimes takes a long time to contribute a change.
For illumos there is also pretty high expectation of backwards compatibility. What unleashed now offers is a place to do more radical changes and quickly iterate on new ideas.
While the small illumos community is already pretty fragmented my personal hope is that in the future work done by both projects keeps getting merged from time to time. Maybe this also serves as a wakeup call to the illumos project to make some improvements to the contribution process itself.
If your highest goal is that someone contributes code to your OS why would anybody use it? Think about htat question and you might find your actual goal. If you can't find a goal, then maybe the existing OSes are sufficient. Choose one, live with it, and develop something else on top that doesn't have a satisfiable alternative yet.
Now I will admit this is going to be tricky. How do you dissuade junk contributions or misfeatures but at the same time encourage contributions? They may end up with an overall declining quality which kills off other goals such as user adoption.
I have contributed a small 1-time change to the Linux kernel. It was relatively easy and straight forward; I sent my first patch to the wrong list, but the maintainers helpfully corrected me.
The difficult part is finding what to contribute, IMO.
And that's just for a small change. Large and/or controversial changes often involve multiple rounds of review and can take (literally) years to get upstream. I'm sometimes amazed when I read LWN about how long it has taken for features to get integrated into the kernel.
I don't know what Unleashed OS is planning but it seems obvious we could make this a whole lot simpler.
Openness in general requires a large learning curve, and I think it's a mistake for any open source project to be marketing itself as a democratic entity that literally anyone can contribute to. If anything, open source is more like a technocratic government (ergo the "Benevolent Dictator For Life" position in a project). Even Wikipedia is a huge pain in the ass to add to at this point.
I've noticed a similar phenomenon with the open science movement, where scientists publicly release their data. Unsurprisingly, few folks from the general public usually work with open data with the necessary assumptions to draw valid conclusions from it. Other scientists, as insiders who've built up decades worth of expertise in statistics and other analytical techniques (at least in theory; in practice not always) are usually the only people who have the ability to fully milk the data for what it's worth. Science, however, is possibly even slower and more political than linux kernel development, although I don't have enough knowledge to make a proper comparison.
These are all pretty well documented tasks. It might not be 'one button simple' like a gitlab merge request, but I thought it was pretty easy overall. It did take some time investment on my part to learn the submission steps (about a weekend total, including the compiling and testing locally).
> Large and/or controversial changes ...
I suspect if they are large or controversial, that in itself implies the barrier to merging is going to be higher. How many new contributors (non-paid contributors especially) want to merge something controversial on the first go?
Could it be simpler? Possibly. Sometimes complex systems do require complex processes for maintenance. I'm just saying I didn't think it was especially burdensome to contribute a small change as a brand-new, completely-unassisted by another person (before sending patch) contributor.
I agree though, that it might be good marketing to claim that. Especially younger people haven't seen that 3-5x already so they still think it might be possible. Just like a 4 Hour Workweek, or that you could be Fluent In 3 Months in a new language.
Is this an operating system for people who don’t care to make a good first impression?