I suppose that with a lot of effort along those lines you could make
an SSH implementation behave like a UDP-encapsulating VPN mechanism. I
haven't read the paper carefully, but I'd expect that you would have
to reimplement a lot of the TCP acking and retransmission stuff for
most of the SSH connection (e.g. you definitely need to retransmit the
KEX packets if they get lost), and only disable it for the special
case of data packets belonging to a special VPN channel type.
But what is the advantage over using dedicated VPN protocols which
have been properly security-reviewed for this use case? The only
practical benefit I can think of is if you wanted to reuse your
existing SSH login credentials, such as a private key stored in your
ssh-agent.
--
Simon Tatham These are my opinions. There are many
<
ana...@pobox.com> like them but these ones are mine.