Open source / free software has always (long before it had those names) had much productive innovation.
And of course it has. Open source is all about mimicking the open science research community around all the universities in the world. This scheme has been perfected for centuries with the explicit goal of more innovation.
With open source you don't have to reinvent the wheel. Have a new idea for an operating system feature? Implement on Linux and you can be ready in no time and (which is important) you have plenty of people to show it for, get feedback, and if it turns out good, plenty of users. Creating a new language? Just write a new frontend to gcc and you have modern state-of-the-art code generation and register allocation tools which would cost you millions to develop from scratch.
Five years ago, there was already a complete operating system for us, GNU/Linux, all created within the open community using open tools. Using only some of the thousands of available tools, you could chat with users around the globe using IRC, setting up ad-hoc document collaboration with CVS and LaTeX, develop software etc.
Corporate users saw there was money to save by using free software, *if only* there was a free word processor. Word processor! You had emacs, CVS and LaTeX and you wanted a *word processor*! Why? Because the better software wasn't "familiar enough". Ok. After several years we have, what, 5 good free wordprocessors? Repeat ad infinitum.
Now we find ourselves in the peculiar situation where new users are introduced to lookalikeware such as GNOME, Evolution and Openoffice -- and they complain they are just pale clones. Well of course they are! That's why they were created!
Want a 3D GUI? Available. Unified messaging? Available (at least with some Perl glue). Ad-hoc collaboration systems? Right there.
If you want to understand the innovation taking place in the free software community, start with understanding our basic tools. Understand CVS, LaTeX, Perl and such. Look at the reiser4 filsystem, the XML-UI in Mozilla, Parrot and Perl 6.
But as with all kinds of science -- not all innovation is good. Actually, very few is. Study a place such as linux-kernel. Most features are not considered for inclusion in the kernel, but a select few are. All this discussion and all these features are important to select the few good ones.
Study free software from a science perspective and then share your thoughts. The article doesn't do the process justice.