question about AS relationship

Song Li refresh.lsong at
Fri Feb 21 11:57:45 UTC 2014

Thanks. I'm doing some research on route leaks, you are a great help to me.

Sky li

> On Friday, February 21, 2014 08:57:07 AM Song Li wrote:
>> the AS relationship between AS1 and AS2/3 is peer, and
>> AS1 cannot announce routes from AS3 to provider1 by
>> rule.
> Or even Peer-AS2's routes to Peer-AS3 (and vice versa), in
> general best practice filtering rules, unless transit is
> requested.
>> But if AS1 do it, and the realtionship between AS1
>> and AS3 is invisible to provider1, how can provider1
>> detect this route leak without knowing the privacy?
> Provider-1 wouldn't care whether it's a route leak or not.
> In Provider-1's mind, Peer-AS3 could (suddenly) be a
> customer of AS1. And since AS1 is a customer of Provider-1,
> Provider-1 will be happy to move those packets along as it
> represents more revenue for Provider-1 (more so if traffic
> is sold on a 95th percentile or volume utilization basis).
> It is, really, up to AS3 to detect that AS1 has leaked its
> routes (or paths, to be precise) to Provider-1, and then
> pick up the phone and scream at AS1 to get that leak fixed
> plugged.
> Of course, all of this is a moot point if Provider-1 is a
> good provider and makes sure they only accept routes and
> paths from AS3 that AS3 should be sending to Provider-1 in
> the first place. But as we know, some providers are a bit
> (actually, very) lazy here.
>> In other words, could the business relationship between
>> AS1 and AS3 be known to provider1/2?
> Not really (or not that easily, to be specific).
> With enough time and access to several looking glasses and
> public route servers, one could "infer" (to a certain degree
> of error) business relationships between peering
> relationships, i.e., whether they relationships are
> customer, peer or provider.
> But in your particular case, unless AS3 has a direct
> connection toward Provider-1/2 (where a route leak would
> introduce more problems), Provider-1/2 don't really care
> about whether this is a leak or not from AS1.
> But again, this whole discussion is mooted if Provider-1/2
> do proper background checks and filtering before they turn-
> up the service for AS1.
> Mark.

Song Li
Room 4-204, FIT Building,
Network Security,
Department of Electronic Engineering,
Tsinghua University, Beijing 100084, China
Tel:( +86) 010-62446440
E-mail: refresh.lsong at

More information about the NANOG mailing list