Brian E Carpenter wrote:

One thing that may be useful is an API that allows the upper layer to
tell the local IP stack "Hey, I'm seeing a liveness problem, please
check it out." That would make constant liveness checks by the network
layer unnecessary.

Right.  This would also account for the fact that the result of probing
does not necessarily correspond to the success/failure of data delivery
-- independent of whether the probing packets are sent by a host or a
router.  One potential cause for incorrect probing results is that data
packets get filtered due to the protocol they carry, whereas probing
packets pass through filters.  Another potential cause is that data
packets and probing packets are routed via different paths (cf. ECMP) of
which only one is functional.

Only the use of data packets themselves as probes can avoid this issue.
But then, if we use data packets as probes, we eliminate the overhead of
probing in the first place...

- Christian


_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg

Reply via email to