Justify TCP replay protection claim

[DemiMarie]

The specification claims that TCP with ISN randomization and PAWS provides good replay protection, but gives no proof.

[insanum]

I think the point being made in the spec is to keep PSP simple (and easy to implemented in hardware), thereby relying on the layers above to handle other security concerns. It's not just TCP... the idea is any protocol could be wrapped with PSP. Outside of TCP, I suspect there will probably be traction with RoCE and/or some kind of lightweight socket-based alternative to QUIC.