said by SYNACK:I wonder why the author decided to keep the source port constant? Implementation of a random source port would have made this even nastier to track.
I was/am wondering that, too, but it occurred to me that it's probably faster to bind to a single port and then fire out 20000 packets than to both grab a new source port (indirectly via the OS) and send out the packet 20000 times. (Notice how I made the former sound faster... )
That's what I was speculating about with "performance" above.
I do agree that letting the OS pick an open ephemeral port would make this harder to track.