Various types of filtering - why not?
Why doesn't the community seem to be able to fix its own biggest problems even when technical solutions are available, or could be made available? Anti-spoofing ingress filtering and route advertisement filtering can be implemented, and it wouldn't be that hard, so why hasn't anyone stepped up and said "do it?" If I'm not mistaken, a few years back, Sean Doran at Sprint and a few other NANOG members took a look at the Internet and said "something is not right." (Sean, I know you still poke in every now and then, so forgive me for putting you on too high or not high enough a pedestal.) The mess of small routes cluttering up everyone's tables was handled with prefix filtering. Flap dampening was implemented and major carriers began to use it, solving the problems of excessive BGP noise. The community said "fix it" and it was fixed. We can thank them for a more stable, cleaner network than we would otherwise be left with. How come now, five years later, nobody's willing to say "fix it?" Come on, everyone, let's stop joking around and put spoofing and bogus announcements behind us so we can move on to whatever the next big problem is. Corporate maneuvering aside, the major NSPs today are more or less the same as the major NSPs of 1995. Have things really changed so much since then that we won't be rid of spoofing and bad routes until someone creates an agency to regulate us? What does that say about what we've become? So I'll say it: We've got these problems staring us in the face, and we've got the know-how and the means to fix them. Whether that means moving straight to implementation, or if we need to strategize for a while, it can be done. Now let's do it. Mark
Randy Bush wrote:
route advertisement filtering can be implemented
this is not true for the more interesting cases. see talk at most recent nanog.
Are you talking about routing registries? RR brokenness doesn't mean that filtering is impossible, it just means that they should be fixed. And they would be fixed if their use was required. Mark
participants (2)
-
Mark Mentovai
-
Randy Bush