We're discussing the merits of particular licenses, particularly their impact on the software world, so this is kind of tautological, don't you think?
In any case, I was not aware of that difference between Apache and BSD, so you have my sincere thanks for pointing that out. I've got a few things on Github to modify methinks...
But people contribute when they've got reasons to do so and many people believe in copyleft. And the GPL is an interesting case, because it happened because its author hated copyright, so he decided to turn it on its head and use it to his advantage. And in a world in which copyright (and other IP laws) keep getting worse, copyleft gets even more relevant.
To the issue of whether copyleft did any good versus MIT/BSD, well, would have Apple open-sourced WebKit if it wasn't for KHTML's LGPL license? I'm inclined towards no, but truth is, we'll never know.
On GCC on the other hand, its authors not only used GPL, but actually made it hard in the actual code for proprietary extensions to happen. This might have been good 15-20 years ago, when the landscape was dominated by proprietary compilers, but now I consider that we're better off with LLVM taking off.
Yes, it does. You set up a pretty big straw man there, allow me to illustrate:
"Smoking is bad"
"I don't know, I have very good friends who smoke."
There is undoubtedly great software that was shipped under the GPL but they became successful because they are great software, not because of the license they use.If encouraging your project's success is secondary to preserving the freedom to scrutinise, modify, or redistribute its code, then GPL should be considered. Windows is successful by many metrics, but a failure where those freedoms are concerned. Obviously this is a good trade-off for Microsoft. It's not a good trade-off for Linux.
Where derivation is concerned, MIT & BSD guarantee diametrically opposed freedoms to the GPL. Which freedoms are more important depends on the intentions of the author, and the nature of the project.
The question here is about the choice of licence affecting the success of free software.
My argument is that, if success is more likely under an MIT style licence, this still may not be enough to sway me if I want a guarantee that users of my software are able to scrutinize the source code, or that derivatives of my software will remain free.
I can put a "hello world" app out there and release it under GPL, but is it useful? Does it actually solve any problems?
In any case, freeness is a goal for me as a user of software. I choose Gimp over Photoshop, for example, because I care more about still being able to access my photos in 20 years time than I do about having access to certain feature sets.
The purely political stance I rail against is, in effect, "X fits my use case and is objectively better in every way, but Y is free, so I'll use Y."
For what it's worth, I can still open PSD's created way back in the 90's, and the format is well documented.[1] Adobe could disappear in a puff of logic tomorrow and their format wouldn't become unreadable.
[1]: https://www.adobe.com/devnet-apps/photoshop/fileformatashtml/I would argue that "hello world" is useful only if you release it under a very permissive non-copyleft free software license or, better, as public domain, for exactly that reason. Copyleft is excessively restrictive for this purpose.
If a software author wants for people to use/distribute his software under a specific license, that's his prerogative.
And the GPL is a truly free and open source license, it does allow you to use it commercially and for whatever reason, it does allow you to create derivates, all it asks is that in case you redistribute a derivate, you must also give the source-code on request. It's not even an EULA, it's just a copyright license.
And btw, compared to BSD/MIT, the GPL also contains a patents grant (implicit but powerful enough for version 2, explicit for version 3). And unfortunately in this universe, the big companies contributing to open-source also have really big patents portfolios, therefore I always look suspiciously at code licensed with BSD/MIT. And unfortunately patent trolls are a reality.
If you prefer more permissive licenses, then go for the Apache 2.0 ;-)