Jim,

> There are two performance questions: cost to establish the connection
> and the cost per packet. Any tunnelling scheme gets both. The
> scheme I proposed integrated the security handshake with the
> existing protocol. Yes, some of our message get longer, but that in
> itself is insignificant -- it's the number of round trips that
> matter.

I must be missing something, but if you write (item 6):

"If an encryption algorith other than None and a session
key is supplied, all bytes following the lpb and all subsequent
packets in either direction are encrypted with the session key."

How does this scheme differ from the tunelling one where each packet is also
encrypted?

Roman