The GPL in practice hasn't always resulted in more freedom compared to other licenses. For example, GCC is purposefully designed to be non-modular, to prevent people from developing proprietary front ends. You may agree or disagree with the trade-off, but observe the LLVM project: you have the freedom to manipulate IR outside the compiler. With GCC, you have permission to manipulate IR outside the compiler, but they've gone out of their way to make it difficult for you, so you might as well give up. LLVM's license makes the idea of impeding proprietary front ends a non-starter.

But that is not a result of the licensing, it is a result of a design decision. While it may very well be that such a design decision was made to avoid GCC being used with proprietary front-ends (along with the fact that GCC is simply an older project), it does not provide an argument against GPL, at best against some choices made by the FSF. And after all, the LLVM license is the reason why people have to hope that Apple will open Swift, as opposed to that being a fait accompli. And the LLVM license is the reason why hardware vendors will be able to build proprietary back-ends. In my book neither of these facts increases freedom.

as much as i dislike GPL, i think that the success of LLVM/Clang comes from being a much superior platform to GCC. having built languages targetting both GCC was frought with "Who on earth implements this like this" moments. LLVM was painless, everything was obvious and the level of obfuscation is minimal... the documentation was helpful occasionally, rather than vitally necessary.

That was Linksys not Netgear.

re: GPL/LGPL being the most successful software licenses, that seems to be changing in favour of permissive licenses like MIT/BSD/Apache

Back in 2013, Aaron Williamson of the Software Freedom Law Center did an analysis of licenses used on Github. The MIT license outnumbered all variations of GPL + LGPL licenses combined, second to MIT on the list was BSD (of course, it should be mentioned that most repositories on Github lacked any identifiable license at all)

http://www.softwarefreedom.org/resources/2013/lcs-slides-aar...

Martin Thoma analyzed PyPi (Python Package Index) metadata in January and also found that no license was the top category, but where there is a license, MIT/BSD/Apache licenses outnumbered GPL/LGPL by quite a bit.

http://martin-thoma.com/analyzing-pypi-metadata/#tocAnchor-1...

That shouldn't be very surprising. I tend to licence small projects under MIT style licences because of the relatively small investment of effort. But if I was starting a much bigger project I would seriously consider something like GPL. It would be interesting to see how the use of licences scales with project size and longevity, rather than pure project count.

Richard Fontana analyzed Debian packages... guess what? Mostly GPL.

There are lies, damn lies, and statistics.

It seems perfectly plausible that debian packages are mostly GPL while github is mostly MIT. (Eg Github first became popular in the ruby/rails community which tends to skew more MIT and corporate-friendly). I don't at all see the implication that the difference means someone's trying to lie with statistics.

Just to clarify, I wasn't trying to imply that anyone was lying.

The lies comes from people who uses the statistic to prove a change in society. The statistics is, as always, just data that is relevant in a specific context.

As to why github is mostly MIT, a reason I read is that github replaced private cvs, ftp and folders on ones laptop with a web service that has the dual functionality of work platform and collaboration platform. It follow the same path when people moved away from using office programs and started to use google docs.

I don't think Richard Fontana did the analysis, in his 2012 presentation entitled "The decline of the GPL and what to do about it" [1], he does refer to an analysis of Debian packages but that was done by John Sullivan (which was prompted by and counter to two analyses showing GPL's decline which were done by Matthew Aslett)

[1] "The decline of the GPL and what to do about it" https://events.linuxfoundation.org/images/stories/pdf/lfcs20...

Cool thanks, you're right it was John Sullivan... should have looked it up before trusting my memory...

I was thinking of his original talk:

http://faif.us/cast/2012/feb/28/0x23/

i'll always concede that these technologies are extremely relevant in the web backend space, but for the most part their are bizarre curios that some technical people know about and even fewer actually use.

gcc has done okay, and its a fantastic project in many ways... some of the optimisations its capable of are really quite smart. that being said though it has usually been held up as example of how hairy compilers can be. there was a period where gcc saw some real use from apple and sony, but other than that its always been considered the outsider in practice... at least during my career. maybe in the dark and distant past when it was even harder to use GNU/Linux and Unices then it compared favorable against borland and the ms vc 5 compiler... but i doubt that is true.

I'm never going to be onside with licenses like GPL when there are MITs and BSDs which do not impose draconian restrictions to help further a philosophy instead of being actually free. The single most common reason I hear not to use a library, borne from practicality, is that it is GPL or LGPL licensed.

The fact that you cite something that, to a close approximation, nobody has heard of as an example of how this stuff has made the world better is a brilliant example of what I am talking about.

Open source is great in lots of ways, but I'm convinced it would be better if its proponents and contributors were a little more in touch with reality.

The single most common reason I hear not to use software is that they are proprietary and demand money up front for using it. Every time torrents and piracy is brought it, software is in the middle of the discussion.

Equally convince that you are about your reality, so am I that opponents to GPL are exclusive those who wish to add restrictions on software. Those who only wish to share software and doesn't add restrictive licenses to their work can treat GPL, MIT and BSD as equivalent. It would be nice if the vocal minority would in this regard be "little more in touch with reality" and be upfront about their intension.