Since this thread started as comparison of the tools, there are two issues 1. Which BGP feeds the tools use? RIPE, RouteViews, other private feeds. 2. How they decide what to send and what not to send? In this case, BGPMon detected an event that was not detected by others, and there might be other hijacks that were local in scope where PHAS or Watchmy might catch something that BGPMon does not. But that does not make one tool better than the other, unless this pattern is repeated. Eventually all tools will catch up with each other on the feeds (or so is the hope), so the difference will then lie in "the decision of what to send and what to drop", and as Todd mentioned, thresholding is very critical here. Mohit Date: Thu, 13 Nov 2008 20:27:32 +0000
From: "Alexander Harrowell" <a.harrowell@gmail.com> Subject: Re: Prefix Hijack Tool Comaprision To: Todd Underwood <todd@renesys.com> Cc: nanog@nanog.org
OK. This seems to be a flaw in RIPE RIS, a pity because BGPlay is great.
- original message - Subject: Re: Prefix Hijack Tool Comaprision From: Todd Underwood <todd@renesys.com> Date: 13/11/2008 8:05 pm
alexander, all,
On Thu, Nov 13, 2008 at 07:56:26PM +0000, Alexander Harrowell wrote:
It may be the North American NOG, but it's been said before that it functions as a GNOG, G for Global. I don't think Brazil is insignificant. I respect Todd's work greatly, but I think he's wrong on this point.
you misread me.
i did not say that brazil was insignificant. it's not. it has some of the fastest growing internet in latin america.
i said that *this* hijacking took place in an insignificant corner of the internet. i mean this AS-map wise rather than geographically. this hijacking didn't even spread beyond one or two ASes, one of whom just happened to be a RIPE RIS peer.
real hijackings leak into dozens or hundreds or thousands of ASNs. they spread far and wide. that's why people carry them out, when they do. this one was stopped in its tracks in a very small portion of one corner of the AS graph.
as such, i don't count it as a hijacking or leak of any great significance and wouldn't want to alert anyone about it. that's why i recommend that prefix hijacking detection systems do thresholding of peers to prevent a single, rogue, unrepresentative peer from reporting a hijacking when none is really happening. others may have a different approach, but without thresholding prefix alert systems can be noisy and more trouble than they are worth.
sorry if it appears that i was denegrating .br . i was not.
t.
-- _____________________________________________________________________ todd underwood +1 603 643 9300 x101 renesys corporation todd@renesys.com http://www.renesys.com/blog