16.03.2011 15:50, eyecue kirjoitti:
Personally I plan to avoid threads (python GIL problems etc) and
instead use stackless python.
But if you are think about using one thread per connection do consider
how it will scale, perhaps better to do it right the hard way.
http://stackoverflow.com/questions/880570/question-about-server-socket-programming-model/880654#880654
This highly depends on the usage scenarios. RPyC isn't the kind of
library that would be used to publish services to Internet. I'd bet that
almost everyone makes only a handful of concurrent connections to each
RPyC server at most.
http://docs.python.org/dev/library/concurrent.futures.html
Is worth a consideration, I think alex is the expert with python
executors in python 2 http://pypi.python.org/pypi/futures.
gevent, eventlet, Coroutines approaches might be worth a look too
gevent/eventlet is out of the question as it requires C extensions and
thus rules out Jython/IronPython and possibly PyPy compatibility. The
problem with asynchronous programming anyway is that you have to
restrict your programming style to fit the model.
Alastair
