🚀 go-pugleaf

Steve,

You ask just the right questions...

A TSB terminates a TCP connection and a TP4 connection.  Hence, it
represents a single point of failure.  Further, the connection terminations
mean that there is no end-to-end checksum.  (A minor additional point is
that this means some extra overhead, since two checksums are computed.)

Just so the point is not lost, note that the primary benefit of the TSB is
that its use, in conjunction with ISODE-like software at the TCP-based
end-user host, means that the underlying IP network AND the kernel
operating system of the end-user's host do not need to be modified.  ALL
of the software can run in the user's application space.  The only
(potential) impact upon the end-user's host manageer is having to
run the application server (responder), if that is a desired service.

The Tunnel operates at the CLNP/IP datagram level.  Since it is stateless,
multiple Tunnels can be made available and data can be alternately
routed, providing robustness against Tunnel failure.

Since it does not participate in the Transport-level ("end-to-end")
mechanism, there is only one checksum (TP4's) and it is legitimately
end-to-end, as it was intended to be.

Just so this point is not lost, either, the Tunnel requires a modified
dual-stack of TCP and OSI in the TCP end-user's host kernel operating
system.  (Note that the modification is to have CLNP talk to IP, rather
than to the link-layer driver.)  However, the network administrator
continues to be able to be unaware of the game.  Only the host is
modified, not the IP backbone, although the Tunnel needs to be added to
the backbone, looking -- as far as the net administrator is concerned -- like
just another IP router.

Dave