On Wed, May 09, 2012 at 11:03:50AM -0400, Jakob Heitz wrote: > You mean for the case when a router does "treat as withdraw" for > routes in an errored update message, but does not bring down > the session, to deprefer the rest of the routes from that peer. > > Right?
Correct. The topic this thread was covering was "what do we do when we receive a very large number of faulty updates". If a session gets a few bad updates, it's reasonable to expect that a few "bad routes" got into the Internet's table. If we get a very large number, it may be safer to assume that something is very broken upstream. When we've decided the upstream is broken enough, for some threshold of broken, two reasonable actions are: - Bring down the peering session. - Exercise "graceful shutdown" on that session to permit its routes to be used as a path of last resort. In this case, we perhaps do not take the session down. The faulty routes are always handled according to the error-handling draft semantics. -- Jeff _______________________________________________ GROW mailing list GROW@ietf.org https://www.ietf.org/mailman/listinfo/grow
