That might be right, because Freshmeat is mostly desktop applications and small utilities.
If you look at more recent projects, especially web-related projects (such as web frameworks) there is an increasing trend towards more permissive licenses. Looking at frameworks (from http://en.wikipedia.org/wiki/List_of_web_application_frameworks):
RubyOnRails: MIT
Django: MIT
CakePHP: MIT
Codeigniter: BSD
Zend Framework: BSD
Symfony: MIT
Turbogears: MIT & LGPL
jQuery: MIT & GPL
Dojo: Academic free
Prototype: MIT
Script.aculo.us: MIT
Y! UI: BSD
All the apache javaEE projects: Apache
Out of the top projects on GitHub, I could only count two that were GPL'd: http://github.com/popular/forked
So out of all of those frameworks, which together cover the vast majority of web applications and web services being built today, not a single one of them is GPL only.
So web projects are very much trending away from the GPL - and the reason why is what I discussed in this thread previously ie. allowing companies who build on top of these products more flexibility. The developers, by choosing a more permissive license, have elected more flexibility for users, and thus potentially more popularity, over a sharing restriction.
That is my point. Those who have no intention of releasing code are not going to regardless of what the license says.
When you have a party that derives code but then becomes uncooperative, what value do they add to the community in any case?
The key benefit of open source is that with an efficient ecosystem of code production and maintenance, developers have a readily accessible repository of code that solves known common problems.
This code can be used so that wheels are not re-invented, and so that businesses become more efficient by instead focusing on 'differential' aspects of their service or product offerings rather than plumbing code.
Since Linux, operating systems are now commoditized, and companies such as Google and Amazon have gone on to use this 'plumbing' to build great services. Without open source, both companies would have spent an extra 5-10 years on development, and likely would never have existed.
The problem now is that the FSF looked at all these companies with billion dollar valuations who were all built on open source plumbing code and decided that it wasn't fair - so they are trying to put an end to it with GPLv3 and the new services provisions.
These new forces being applied to what is an extremely efficient ecosystem of code sharing will completely break apart open source as we know it. The next Google or Amazon will have to look at either other commercial solutions, or more liberally licensed plumbing code in order to build their services platforms.
Forced sharing is simply another restrictive force in the economics of open code. Companies will be a lot less likely to contribute and a lot less likely to participate in this environment if they are restricted in what they can do with the code they have built on open code.
I can name dozens of examples, and more recently they have mostly been with more liberal licenses.
The key question is, does enforcing sharing in the license actually provide for a better overall open source ecosystem.
From my experience, the answer is no - because the downside of having companies not wanting to touch GPL'd code for fear of legal challenge and problems *FAR* outweighs the potential benefits of having those companies contribute resources with more liberal licensed code.
Overall, I find that companies are *more likely* to contribute or to allow developers who work for them to contribute code if the license is a simple, more liberal license.
ie. the parameters of a BSD or MIT license are much easier to understand, and the decision making within a company is easier - especially considering that the driving force for wanting to share code usually comes from the bottom-up, ie. from developers themselves.
The theoretical question here is - would the top 10 GPL'd projects have been more or less successful if they were released under a more liberal open source license that didn't enforce sharing?
I want the people who extend my software to give their extensions to the world to share, the same way I gave them my original program. So, my payback for writing Open Source is that my software drives a further increase in the amount of available Open Source software, beyond my individual contribution.
Has anybody ever proven this?
ie. has it ever been proven that attaching a 'must share' clause to a license (ie. GPL vs BSD) actually results in more people sharing code.
I am inclined to think and believe, based on experience, that it does not. Those who share are likely to share regardless of license, ditto to those who take your code and improve it with no intention of sharing.
Just how much does 'sharing' contribute to open source anyway, considering that all the top projects are tightly controlled by a small number of lead developers who hold the keys to commitments and in accepting patches. Code being shared will likely just go unnoticed anyway.
So, after 10 years, has anyone proven that the GPL works?
No amount of careful planning will ever replace dumb luck.