Sprint does indeed dampen customer paths, and often. On Fri, 21 Feb 1997, Jim Rowh wrote:
The degree of prepending you show is probably more for Sprint's benefit than anyone else's. 1793, 1239, and 1790 are all Sprint AS's. If you're lucky, you can find someone familiar with Sprint's confederated routing scheme to explain the details. AS3505 belongs to VNet, who uses both Sprint and Gridnet as providers. Prepending allows them (VNet) to effectively designate Gridnet as their primary (all things being equal, better--shorter--AS path wins).
As for hold-down routes (to Null0) and damping, yes, they're both good and generally practiced ideas. As to whether and to what extent Sprint uses either or both of them, you will have to take that up with them: noc@sprint.net or 1-800-230-5108
jim
Sorry for littering NANOG with this email, but I'm hoping for a little education on this matter from some of the larger National Service Providers.
This morning we had a customer call in concerning a network hole. Upon investigation, I found the following..
BGP routing table entry for 166.82.0.0/16, version 541808 Paths: (1 available, best #1, advertised over IBGP, EBGP) 1793 1239 1790 3505 3505 144.228.159.17 from 144.228.159.17 (144.228.159.1) Origin IGP, metric 0, valid, external, best Community: 1 Dampinfo: penalty 656, flapped 18 times in 02:19:29
Two immediate things came to mind here.
1. This is a /16 that is flapping. 2. The end user is using AS path prepending for some reason.
With these two facts in mind, the following questions arise.
Anchoring or tying down of a route with a static route to null seems to be a very basic idea (far more basic that AS path prepending), so why isn't this route tied down?
I only see one route to this prefix, why are they prepending? (I admit that only the user of this prefix can answer this question, and I also admit that with 53 flaps on the books so far, the other provider may be dampening the route for us *thank you*)
Finally, does Sprint not apply *any* dampening to customer routes?
My cat is long dead (10 years at least), so it's safe from my curiosity.
Chris A. Icide Nap.Net, L.L.C.