As part of Facebook's open source initiative, it has open sourced a core piece of FriendFeed, which it announced it was acquiring only a month ago. Dubbed Tornado Web Server, it's "a non-blocking web server" and collection of tools written in Python and designed for high levels of scalability. Facebook Director of Products Bret Taylor has a blog post up about it, and says it includes a complete array of site building blocks, including templates, signed cookies, user authentication, localization, aggressive static file caching, cross-site request forgery protection, and third party authentication. There is complete documentation here, and a link for downloading Tornado. Here are more details.
As David Recordan notes on the Facebook Developers blog:
"FriendFeed, which we recently acquired, built their entire site to support real-time updates...[It can} handle thousands of simultaneous connections, making it ideal for real-time Web services."
Recordan also supplies the following chart showing how speedy Tornado is compared to existing web frameworks written in Python (Django, Google's webapp, web.py):
 Recordan adds:
"It is no longer just the traditional Linux, Apache, MySQL, and PHP stack that make a site like Facebook or FriendFeed possible, but new infrastructure tools like Tornado, Cassandra, Hive (built on top of Hadoop), memcache, Scribe, Thrift, and others...we believe in releasing generically useful infrastructure components as open source as a way to increase innovation across the Web."
Facebook continues to make substantial contributions to open source like this one, but it also remains on the heat seat for the underpinnings of its own site. Just this week, a Delaware Court ordered the company to turn over all of its source code to Leader Technologies, due to alleged patent infringement. With 250 million users and counting, the company has its own internal development challenges, but it's good to see Facebook giving back to open source.