On Tue, 26 June 2001, Andrew Partan wrote:

On Tue, Jun 26, 2001 at 11:48:29AM -0700, Sean Donelan wrote:

...

RESOLVED, the error handling mechanism should be revised to reject only the portion of the announcement in dispute instead of disrupting the entire BGP session.

The big problem here is that it may be impossible to tell that just this bit of <goo> is bad and that everything received after that is fine.

I'm willing to accept there are cases you can not recover, and the only acceptable alternative is to reboot (gee, Microsoft thinks that is the solution to everything). My objection is RFC 1771 does not give implementations the ability to continue when they can. Aborting the BGP session should be an option, and extreme option, but not the ONLY option. So should throwing away the bad data, alerting, and continuing with good data.