After an ack on IRC, I have now pushed the srbchannel patches to master. The new protocol is enabled by default. In short, the new protocol uses a shared ringbuffer, which is faster on low-latency workloads because the per-packet overhead is lower, compared to the unix socket/pipes we use today. You should notice a difference on both server and client side. It takes a packet extra to set up so extremely short connections (e g pactl) might take slightly longer to execute, but you're unlikely to notice. If you want to disable the srbchannel, you can add srbchannel=0 to protocol-native, like this: load-module module-native-protocol-unix srbchannel=0 Enjoy, and report if you find any issues. Also thanks to Peter Meerwald for helping out with benchmarking, bug finding and testing! -- David Henningsson, Canonical Ltd. https://launchpad.net/~diwic
