[sureglymop]

Quite interesting! I get similar results on my machine. But let's also measure latency which is what should show the protocol stack overhead:

    $ netperf -H 127.0.0.1 -t TCP_RR
    Local /Remote
    Socket Size   Request  Resp.   Elapsed  Trans.
    Send   Recv   Size     Size    Time     Rate
    bytes  Bytes  bytes    bytes   secs.    per sec
    
    16384  131072 1        1       10.00    63506.58
    16384  131072
    
    $ netperf -t STREAM_RR
    Local /Remote
    Socket Size   Request  Resp.   Elapsed  Trans.
    Send   Recv   Size     Size    Time     Rate
    bytes  Bytes  bytes    bytes   secs.    per sec
    
    212992 212992 1        1       10.00    122526.23
    4608   2304

But I would agree now that it probably doesn't matter for most applications.