HTTP connection pooling #133
Description
A customer has reported that when doing this they get expected reuse of the HTTP connection pooling:
client = AblyRest('api:key')
channel = client.channels.get('channel_name')
channel.publish('foo', 'bar')
# Slower request expected as new TCP TLS connection needed
channel.publish('foo', 'bar')
# Faster as previous connection reused automaticallyHowever, when they do this, they don't get HTTP connection reuse:
client = AblyRest('api:key')
channel = client.channels.get('channel_name')
channel.publish('foo', 'bar')
# Slower request expected as new TCP TLS connection needed
client2 = AblyRest('api:key')
channel2 = client2.channels.get('channel_name')
channel2.publish('foo', 'bar')
# Equally slow request as firstIn other libraries, the underlying HTTP library automatically handles connection reuse so long as the requests are in the same process, whereas with Python it seems this is not the case and the pooling is only done with the instance REST client.
This is not the intended behaviour.