[brunoborges]

It's not the same thing, because of the flow:

- developer finds a cool project - developer consumes the cool project - developer finds opportunity for bug fix or enhancement - developer forks and starts working on the enhancements - developer spends hours/days writing code - developer submits a PR - developer gets a notification (upon filling the PR, or after, from a bot) - developer is now extremely frustrated - developer complains - maintainers and the developer start debating - the Pull Request storm goes viral - developer threatens to stop using cool project - maintainers handle the situation poorly (they are coders, not Public Relations people) - the community makes things worse

And so on...

[jrochkind1]

If issues/PRs were limited to collaborators, wouldn't they not be able to submit a PR at all?

[brunoborges]

Here's how I propose a better PR flow in GitHub:

https://news.ycombinator.com/item?id=30705581

[jenscow]

Perhaps the developer should have discussed their ideas before investing their time, if they're going to be upset about their unsolicited code being rejected.

[rectang]

The maintainer still has to deal with the negative interaction even if in some cosmic sense the contributor could have known better. A workflow design which ensures that such negative interactions take place is structurally flawed.

[chii]

what's so negative about pressing the reject button of the PR? The maintainer doesn't have to even spend more than a second rejecting any contributions they get. They don't have to read any of the comments, nor reply to any concerns from the contributors.

The contributors aren't entitled to the time of the maintainers.

[bachmeier]

I don't understand. You have "developer forks". Everything after that is irrelevant. That's the whole point of open source.

[kccqzy]

On GitHub a fork is a lightweight thing. People fork the repo in order to contribute because they can't push to the original repo. It's not like you are philosophically against Emacs and decide to fork it to produce XEmacs.

[bachmeier]

Just as I don't understand the original comment, I also don't understand yours. You don't need to push to the original repo because open source allows you to make your fork available to others. The developer of the original has a right to decide to accept your changes or not, so I fail to see why it matters that your PR isn't accepted.

[tremon]

Just as I don't understand the original comment, I also don't understand yours

The way Github e.a. use forks dilutes the larger meaning of the word, that's what this thread is about. When people used to talk about forks, they always mean a community fracture over differences of opinion: gcc vs egcs, xfree86 vs xorg, ffmpeg vs libav, openwrt vs lede, glibc vs eglibc, kde4 vs trinity, gnome3 vs cinnamon vs mate.

The common-use term of "fork" on github has nothing to do with divergence of development, it's just a band-aid for lack of contributor access control. I can understand why people don't like github's use of forks, and that has nothing to do with what the license "allows" or not: if it's not a divergent development line, it's not a fork, it's a clone at best. In most cases, I'd say it's just a feature branch hosted in a different repository.

Calling something a fork implies long-term viability (or at least the intention) as an alternative to the original repo. That doesn't sound like a realistic description of most cloned repo's on Github to me.

[cxr]

> The common-use term of "fork" on github [i]s just a band-aid for lack of contributor access control

More accurately: it's a consequence of GitHub's early decision to reject the traditional currency of open source (patches) in favor of imposing dark pattern-infested workflows on hosted projects and their contributors.

When you accept patches, anyone can easily create an account, submit their patch, and then be on their way. GitHub's opting for an unnecessarily heavyweight pull request workflow instead, though, meant that contributors were required to set up a personal on-GitHub copy of the desired repo, point their copy on their local machine at that as a remote, and push to it. At that point, they're already paying the cost of project hosting overheads just for contributing to a project that isn't theirs. When it comes time to start their own project, or evaluate whether it's worth it to continue hosting their existing projects elsewhere, pressure to choose GitHub cones into play that wouldn't otherwise if there were a level playing field. Fast forward nearly 15 years of network effects, and you have it as a nearly unavoidable multi-billion dollar behemoth.

It's such a transparent growth hack inspired by the underhanded tactics used by social networks that folks' unwillingness to even acknowledge its negative impact on productivity and privacy is really infuriating—not to mention the negative impact itself.

[theobeers]

“Feature branch hosted in a separate repository” is a very good way of putting it.

[jlokier]

> I fail to see why it matters that your PR isn't accepted.

Making their code useful to other people is a strong motivator for many altruistic open source developers.

The purpose of getting PRs accepted upstream is so that your changes are useful in concert with other people's changes, especially their later changes.

Your private change is of little use to others when it's only in your private fork, which hardly anyone knows about, and if they do know, your change is incompatible with later work and maintenance by other people anyway.

If you don't care whether anyone else uses your work, especially in future, that's fine. But if you want it to be more useful to others, getting it merged upstream makes a huge difference to that.

(Merging upstream or just submitting PRs that don't get merged also helps personal visibility and reputation, which is undoubtedly a motivator for many. That's a burden on upstream maintainers when someone is pushing low quality work and doesn't care to ensure it's useful to the project.)

[Nadya]

Maintaining a fork and maintaining a single feature of a project are two different levels of commitment and you can want to do one without doing the other although all too often people want to contribute a feature and then not provide support for it and expect the project maintainer to now maintain the contributed code. But that's a different issue (and why many maintainers sometimes don't accept random PR's even if the code itself is fine).

> You don't need to push to the original repo because open source allows you to make your fork available to others.

This is simply not true - not all open source software is copyleft. You can have restrictive licensing regarding use and/or distribution and still be open source.

[bachmeier]

> This is simply not true - not all open source software is copyleft. You can have restrictive licensing regarding use and/or distribution and still be open source.

Well, of course you can adopt any definition of "open source" you want, but I'm using the OSI definition, which states:

"The license must allow modifications and derived works, and must allow them to be distributed under the same terms as the license of the original software."

You are referring to "source available", for which they offer the clarification:

"Open source doesn't just mean access to the source code."

[Nadya]

There is a reason FLOSS exists as a term to differentiate itself from OSS and I think the ideological differences between the two are important. So does Bruce Perens, the co-founder of OSI and the author of the Debian Social Contract of which the OSI definition was based off: https://web.archive.org/web/20140716055445/https://lists.deb...

This was back in 1999 - it's been over 20 years and "OSS" has only been muddied further and further to mean "source available" since.

Pedantry aside tremon's response sums it up well: "Feature branch hosted in a separate repository."

[jrochkind1]

Y'all are talking about possibly three different things.

"Source available" does not mean anything about a license at all. It literally just means the source is available... to someone.

Open source generally means it must have a license that allows modifications and derived works. It is more than "source available" it means that a public license to use and redistribute the software is available. "source available" does not necessarily mean you have a free license to use or distribute the software, or modify it and distribute the modifications. "open source" does. The Open Source Institute has tried to standardize a definition. https://opensource.org/osd

"copyleft" is sometimes used to mean something more than some "open source". Richard Stallman has promoted these sense of "copyleft". Like GNU license as opposed to apache. Both are open source -- nobody says apache license isn't open source right? Open source licenses that aren't "copyleft" are sometimes called "permissive". Here's one description of it that will suffice:

> 3. Is the Apache considered copyleft?

> Copyleft licenses require the derivative works or modified versions of existing software to be released under the same license. The Apache License doesn’t have any such requirements. It’s a permissive license. It permits you to release the modified parts of the code under any license of your choice. However, you are required to release all the unmodified parts of the software under the same license (the Apache License).

--https://www.whitesourcesoftware.com/resources/blog/top-10-ap...

So, no, "source available" is definitely not the same thing as "open source" -- all open source may be "source available", but things described as just "source available" usually are not open source. The source might be available to you, but you might not be allowed to use it (at all, or in certain ways) without paying, and you might not be allowed to redistribute derived works.

And "copyleft" is usually used as a subset of "open source", for restrictive/"viral" licenses.

I have no idea what any of this has to do with where this discussion began -- but yes, anything that's open source, copyleft or not, gives you the right to make a variaton of the software, use that variation, and distribute it. I think that is part of the accepted definition of open source. "Source available" software may not give you this right.