>> Because you could see this and mark your /24 not to be bounded (if we >> put that back in). We were trying for the simplest solution to >> increasing table sizes --I know there are economic reasons the table >> sizes are growing, but we need at least some option to de-escalate the >> madness at some point. > > People have been saying for the last twentysomething years that the sky is > falling due to DMZ growth. Geoff Huston gave a talk about this last year > in Vienna where he presented some data and analysis: > > http://ripe63.ripe.net/archives/video/178 > http://ripe63.ripe.net/presentations/61-2011-10-31-bgp2011.pdf
I would suggest: 1. IPv6, over time, is going to present a different issue... 2. That the problem is an economic problem, not a technical one, and hence we need a solution that pushes the tradeoffs between the cost of advertising and the cost of not advertising into the front of our minds, rather than making the advertisement of a new longer prefix route "free." > I'm not suggesting that the DFZ couldn't do with a bit of slimming, but the > bgp RIB is not the best place to do it because you're losing too much > information too early. This isn't at the BGP RIB, this is removing routes from the system when they are no longer useful for modifying forwarding direction. It's no different than removing routes that have lost in the best path decision process, in terms of "lost information." We could, of course, back up and argue that we shouldn't be removing routes that lose in the bestpath process, because we're "losing AS Path information," if we want to argue for complete information everywhere. :-) Russ _______________________________________________ GROW mailing list [email protected] https://www.ietf.org/mailman/listinfo/grow
