> "Since a non-trivial number of network elements discard well-known > ICMP packets the results of our tests do not offer hope for protocol > designers proposing to use new ICMP messages to signal various > network path properties to end systems (e.g., for explicit corruption > notification, handoff or outage notification, etc.)." Yes, this is a very serious issue. Very often, a server that is configured for a service will open exactly the port necessary for that service, and drop all other traffic. For example, a web server will accept TCP packets sent to ports 80 and 443, but it will either drop everything else, or be located behind a firewall that drops everything else. This restriction affects the way we design protocol extensions. I see that as an argument for "in-band" signaling, e.g. parameters in TCP packets or in IP headers of TCP packets, by opposition to "out of band", e.g. ICMP messages. -- Christian Huitema _______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf