Private message from Jens-U


sorry for the pm - I’m not sure if you’d like to discuss this publicly:

If on PC1 I manually do „route add via dev eth0“ then all is fine, but this solution is not applicable in our case.

I’ve recently seen more and more network setups that rely on ICMP redirects - you’re my first to state that explicit routes are a no-go. Would you mind telling me why such routes are not applicable?

AFAICT, ICMP redirects were meant as a mechanism to react to misconfigured networks, and I’m old-school enough to have seen more than my fair share of cases where these messages were blocked on the sender side, didn’t get through (those are just ICMPs, after all) or the receiver wasn’t handling them too well (like with ICMP redirs and that recent Linux bug). So I’d always opt to change the network design or adopt dynamic routing protocols.

I’d like to learn under which circumstances those options aren’t applicable, just to extent my knowledge.

With regards,