That is always touted by GPL advocates, but where is the proof? Where are all the companies enhancing non-GPL-licensed software and not giving anything back? Where are the companies making money selling private forks of enhanced FreeBSD, NetBSD, OpenBSD, Ruby, Python, Ruby on Rails, X11, Apache, etc.?
It doesn't make sense. Maintaining your own changes in a big software project is a pain in the ass. It's worth contributing changes upstream if only so you don't have to maintain your own patches and re-merge them every release.
They're doing it to GPL software. Take a look at pretty much any embedded Linux project and you will find a GPL violation. For example, most ARM devices are unusable without binary drivers that rely on specific builds of specific Linux kernel versions (example - you can't get sound or accelerated video on the Raspberry Pi!). I've had one plug computer maker tell me they have a custom version of Linux and they won't provide the source. Whenever I come across anything like that I report it at http://gpl-violations.org/
Sun Microsystems?
I think part of it has to do with a tipping point: if you can hire up a good chunk of the key contributors and run with it, then it's easy to make a proprietary fork, and at the same time weaken the open source project. The GPL prevents this. The other thing that prevents it is if it's impossible/difficult to hire up enough people for a proprietary fork to outstrip development of an open fork. So with permissive licensed projects, you want a big core team comprising people with diverse interests and incentives.
Also, I think things have changed over time - perhaps, 'back in the day', there was more the idea of not contributing back by default, whereas I think far more companies have made their peace with open source these days, and understanding that being involved and contributing are good things. This tilts things in favor of permissive licenses.
+ Apple
They contribute back only whatever they see fit, and that's in perfect accordance with the licence.
Apple has also contributed a lot that they weren't required to contribute. Sure they are an example, but they are not nearly as cut throat as people like to make them out to be. Look for Apple.com on commits to the FreeBSD source tree...
No, it doesn't. Those key contributors are still the copyright holders of the code they contributed, and they are free to redistribute those contributions under different terms. If they all work for one company, they can just license copies of their code to that new company under non-GPL terms. The GPL fork still exists, but now a new, closed-source version of the code can be made with all of that (now-)employee-written code. If there are enough of those components, that company can just rewrite the code contributed by others that were only licensed under GPL, and now they have a mostly duplicate version of the same program, able to be extended and resold without source code.
Digium does this for the Asterisk project. They have a GPL version that everyone is free to use, but nobody else is allowed to redistribute it without the source code. However, they make all contributors to the GPL project license their code to Digium so that Digium more-or-less owns the code, which Digium then turns around and sells under a closed-source license to companies that don't want to give their changes back to Asterisk.
[1] https://www.microsoft.com/download/en/details.aspx?id=274
Building xnu from public sources is quite an undertaking. It does still seem to be possible (although it got a lot harder with the x86 transition due to more proprietary drivers). The PureDarwin distribution still seems to be based on the Snow Leopard kernel, I'm not sure why. http://www.puredarwin.org/
Meanwhile components like libraries, frameworks, have since long leaned towards permissive or atleast LGPL style licencing due to them constituting only a small part of whatever program they are used in.
Like the edited article states, the author was wrong in saying GPL is declining or that projects are in any way switching, the thing is that these days we see lots of new open source developments in web technology frameworks and these frameworks are predominantly permissively licenced as they are basically components on which full web solutions are built.
And given the sheer number of web oriented open source tech popping up these days it's no wonder we are seeing new permissively licenced code dwarfing new GPL licenced code these days. For example when I viewed the 'Popular Watched Repositories' list on github, the only ones I know are GPL licenced were Linux and Git, and the list was heavily dominated by web targeted projects.
Now there seems to be a strong anti-GPL sentiment in web dev, which is likely due to people working there relying heavily on open source components for their proprietary solutions and as such they simply can't use GPL, hence complaints.
Bottom line is that there is no 'one type fits all' licence, which is why we have licence proliferation. I've always held the belief that whatever the author chooses is the right option, be it permissive, GPL or even keeping it proprietary.
As a case example: Apple has modifications to their version of WebCore (lower-level than WebKit, for context), a project shipped under LGPL (unlike WebKit, which is distributed under a modified BSD license).
However, they don't want to have to give out those changes, so they provide the source code to a WebKit with some minimal changes to call into new code that they then don't provide: it will compile, but you will get link errors.
Understand: this new code may be critical to the functionality of the library, and may be modifying the behavior of or even outright replacing key APIs and mechanisms of the LGPL library, in this case WebKit.
They then provide binary object files for the new functionality that they added, and the result is compiled together to form the unified WebCore dynamic library. Even though the result is a single linked-together library, this is totally legal under the LGPL, as I can still modify the original WebKit code if I wish and still manage to build Apple's modified WebKit binary.
In essence, the boundary around the LGPL code can be made very tight indeed, such that even code that is semantically required for the library to function does not fall under the redistribution requirements of the GPL part of the LGPL code.
The USL lawsuit is what hampered the BSDs early on, and allowed Linux to get so much traction in mindshare, not the license.
However, if the result of my work will be made available under a more permissive license i (most likely) demand less money since it ensures that i can reuse whatever i wrote.
The lawsuit was the big problem.
GPL makes a lot of sense for the former. If you write the Linux kernel, a ton of companies would take it, enhance it, and not contribute back without GPL. That's a big part of what killed BSD. On the other hand, almost no one is going to reimplement Linux just to get around the GPL -- it's cheaper and easier to just give back.
For the little libraries, permissive makes more sense. If your JavaScript control has a license that isn't permitted in my project, I'll just write my own in a day or two. If I enhance it, that's not a core advantage to my business, so I'll likely contribute it back.