A ringbuffer for requests and a ringbuffer for responses would be interesting. Similar to io_uring. The only issue is that the current protocol has variable sized packets which would not work with the ringbuffer implementations.
If the protocol could support data sizes, perhaps with some other mechanism to transmit the packet body, the headers could be zero copy, but some care would be needed on the consumer end to block until it's done working with the header. Probably not really worth it for how light the headers are.
A ringbuffer for requests and a ringbuffer for responses would be interesting. Similar to io_uring. The only issue is that the current protocol has variable sized packets which would not work with the ringbuffer implementations.
If the protocol could support data sizes, perhaps with some other mechanism to transmit the packet body, the headers could be zero copy, but some care would be needed on the consumer end to block until it's done working with the header. Probably not really worth it for how light the headers are.