Ronald van Elburg <vanelburg30@hetnet.nl> added the comment: Problem still exists in 3.7 documentation. It is unclear to me, even after inspecting issues and commits what proper usage of multiprocessing.reduction is. In case it is now fully internal to the module it would be reassuring to read that in the documentation. My code seems to be running without explicit reference to the multiprocessing.reduction module. Which suggest that reduction and rising of connections is no longer needed or handled by the module internally. For python 3.7 I was able to remove the following python 2.7 code: '''python def __reduceConnection(self): # Reduce the connection object to enable sending it if(sys.platform=='win32'): self.connection = reduction.reduce_pipe_connection(self.connection) else: self.connection = reduction.reduce_connection(self.connection) def riseConnection(self,logger): if self.connectionReduced: logger.info('Connection for subscription to {0} with channel name {1} has risen!'.format(self.senderKey, self.receiverKey)) red_conn = self.connection self.connection = red_conn[0](*red_conn[1]) self.connectionReduced=False else: logger.info('Connection to {0} with channel name {1} had risen already!'.format(self.senderKey, self.receiverKey)) ''' ---------- nosy: +Rajve versions: +Python 3.7 -Python 3.4 _______________________________________ Python tracker <report@bugs.python.org> <https://bugs.python.org/issue18837> _______________________________________