David Beazley – Python Concurrency from the Ground Up: Live – PyCon 2015

Does anyone know the details behind what happens to the thread-based server at 11:30? The response time for small requests "drops off a cliff" due to a single large request, apparently due to "the implementation of the GIL"?

Compared to the simple thread-based server, the generator-based server presented in the second half of the talk avoids the above problem. Does the generator-based approach have any other advantages?