On Jan 3, 2013, at 5:01 PM, Jeff Wheeler wrote: > On Thu, Jan 3, 2013 at 4:28 PM, Jared Mauch <[email protected]> wrote: >> Also, if you're seeing some problem, I'm sure that some other set of >> people out there will be seeing it as well. The shared pain/cost will >> exist. If you can't take the risk of speaking BGP, have your ISP send you >> default (or nothing) and advert your prefixes and call it a day. > > There may be no shared pain if you are using BGP only within your > datacenter network. This is precisely why I provided the RainbowPoop > scenario. This is not your 1990s BGP anymore. Anyone who has MPLS > VPNs signaled using BGP already knows this but maybe isn't cognizant > of it in this context. > > You continue to make this argument that breaking your network so you > notice a problem is good. I would like to have the option to notice > the problem by actually parsing my syslogs and NOT having my network > break and cost me and customers a bunch of money. That's not what I > would really do, though. I would just re-actively configure > ignore-bad-messages if I needed to. Because I agree with you, I would > like BGP to flap if something unexpected happens. However, I might > need to stop it from flapping to remain in business.
Oh, I understand all these use-cases, but there is a case for a well-designed network not always sharing/mixing the NLRI. eg: We don't transport v4 NLRI in v6 transport, nor v6 NLRI in v4 transport. If you have a single session with massive shared fate, perhaps it's not a protocol design error but a network design error. - Jared _______________________________________________ GROW mailing list [email protected] https://www.ietf.org/mailman/listinfo/grow
