On Oct 21, 2021, at 5:40 AM, Tomas Prybil <tomas at
prybil.se> wrote:
Here is a small tcpdump with the ping from A34RTR (34.1023) and FORGER (34.101) for
anyone to dig into.
Full pcap file added as attachment.
10:29:02.072710 34.1023 > 34.101 100 link-service 35839>34917 seg 0 dat seg count
0
10:29:02.082747 34.101 > 34.1023 28 disconn-confirm 34917>35839 no link terminate
...
Thanks, that's what I thought I saw at some point in the past. So that's a type
of "ping" that requires only NSP layer behavior; it sends a message to a
connection number that doesn't match a currently open connection, so the other end
rejects it with a "no link terminate" message. That answer is treated as the
ping response.
"traceroute" was mentioned. That could be done, at least a rough approximation,
in a similar way using defined routing and NSP layer mechanisms. It would involve sending
that message with "request return to sender" set in the routing header, and a
visit count that counts down from the limit. That's not a precise solution because
the message will be returned to sender by a router whose "max visits" is less
than what's in the message header. That value can vary, so some of the routers on the
path might be missed. IP avoids this by counting down "time to live" to zero,
rather than counting up "visits" to a node-dependent limit.
I'll put it on the PyDECnet wish list.
paul