I know this has been discussed in the past, but I wanted to comment on a use case that would benefit a great deal from HAProxy->Server piplining feature.
There are many parts of the world where latency is high enough that the initial tcp handshake is a large percentage of the response time. Ideally you would place servers close to those populations, but this gets difficult not to mention expensive. Instead of moving large parts of the serving infrastructure to a remote part of the world, one could deploy a few HAproxies to the region which would then accept tcp connections quickly, and then send along the request via already established connections to a set of back end servers in a datacenter which is farther away. This reduces the latency fairly dramatically. All thoughts are welcome. Mike
