While Facebook and Google are constantly one-upping each other with their open data center designs and open server architectures, Twitter contributes project after project that are applicable to a far larger group of people. Very much appreciated.
I don't mean to put down Twitter's contribution to open source – which is substantial – but I think you're underestimating Google and Facebook's contributions when you suggest that they are only "one-upping each other with ... open data center designs".
I'm more familiar with Google's contributions, but I would imagine both high-user-visibility projects like Android and Chromium and more developer-oriented projects like protobuf, the Closure Compiler, the Dart and Go programming languages, googletest, Guava, Guice, WebM/WebP, V8, Breakpad qualify. I could consider all of these "projects that are applicable to a large group of people". This, on top of large contributions to other key projects like the Linux kernel and LLVM.
I know less about Facebook's contributions, but I do know that their open source projects includes HipHop/HHVM, and they contribute significantly to Hadoop, LLVM and Mercurial (among other high-profile projects).
I think the point here is not on the amount of contributions, it is about: can you build a Twitter clone using exactly the same opensource technologies as Twitter, or how easy is it? Twitter mainly use open source technologies and they will contribute back if they created something useful - even the projects are the core components of Twitter.
For Google, it would be much harder, as they use so many proprietary (backend) technologies, GFS, BigTable, Chubby etc.
