You can clone with
HTTPS or Subversion.
Take out hack to fix logging as it's fixed in Tornado.
* put coroutine running logic in own class
* do as little of the configuration as possible on a per request basis
* remove all per request closures
Don't access IOLoop.instance() until the last possible second.
This lets Tornado HTTPServer's automatic forking work properly.
Include our own version of isgenerator() for Python 2.5 and below.
Don't double-finish Web requests.
Fix a serious bug that would cause the first work item to be repeated…
… on every yield.
Adding a link to the website in the readme.
Adding setup infrastructure.
Adding some (crappy) docstrings.
Adding an initial implementation of Swirl.
Ignoring .pyc files.
Adding an initial readme, and off we go.