Bug Transport pipeline is wasting sending queue!

This transport library have driven me mad for a long time. And finally I spent a lot of time to figure out, what is payloadSize, sendingQueueSize, why sending data return -5 or -4.

The INetworkInterface(typically BaselibNetworkInterface) use sendingQueue * payloadSize to create a buffer, when some code call BeginSend, it create a slice from buffer with size = payloadSize, So maximum BeginSend appending should be queueSize.

Here for the fragmente stage. It also need a payloadSize parameter, which is the max data size the stage accept, typically 32 * MTU.

Here comes the funny thing. When you call Driver.EndSend, internally it use a NetworkPipelineProcessor Sending method. It iterate over all stages of the pipeline, for a single fragments stage, it call Driver.Send(without pipeline) 32 times! So your queue will be used 32 times faster.

And you can't set INetworkInterface.payloadSize = MTU and Fragments.payloadSize = 32 * MTU, it's not allowed. It has to be 32 * MTU.

The modification of package is beyond my capability and time. So could anyone give a solution to me?