At 8:04 PM -0400 5/27/02, Nigel Taylor wrote: >Peter, > It would seem that Cable&Wireless and Above along with RIPE are the >main culprits. > >It would seem to me that this inconsistent route issue would present >problems, what I'm I missing? It maybe that I'm not totally >clear on what constitutes an "inconsistent route". RFC 1930 clearly states >that "one-prefix, one originating AS". I know it's been >mentioned in this thread and I see it noted that the RSNG Project will >notify peers of inconsistent policies registered in the IRR.
I'm not sure I could point you to anything more specific than the IDR, NANOG, and RIPE routing group archives. I hadn't noticed this desire of the RSNG; the impression I have was the inconsistent routes to be reported were those who were NOT registered in the IRR. Such unregistered routes are far more likely to be due to error. RFC 1930, while a wonderful document certainly worth reading by any CCIE candidate, is informational rather than standards-track. >So, how effective is this initiative if most of the community feels it's not >something to be worried about. > >Anyone care to point me in a specific direction. > >thanks >Nigel > >----- Original Message ----- >From: "Peter van Oene" >To: >Sent: Monday, May 27, 2002 6:31 PM >Subject: Re: BGP addressing..i think i understand but i am not sure >[7:45169] > > >> quick comment in line. >> >> At 04:53 PM 5/27/2002 -0400, Chuck wrote: >> >I have a question, Howard - in line: >> > >> > >> >""Howard C. Berkowitz"" wrote in message >> >[EMAIL PROTECTED]">news:[EMAIL PROTECTED]... >> > > This is one of those posts where the attributions have gotten very >> > > confused. Comments inline. >> > > >> >snip for brevity >> > > >> > > It can be done, if both ISPs agree to it and coordinate their routing >> > > policies. A public AS, however, is justified in this circumstance. >> > > >> > > While doesn't quite describe this situation, look at RFC 2270 for the >> > > general strategy. Both ISPs have to remove private AS. This will >> > > also cause more than one ISP to appear to originate the route, which >> > > is a technical violation of BGP (i.e., it's an "inconsistent route"), >> > > but that isn't that uncommon and doesn't seem to break anything. >> > > >> > >> >Question: in an ideal world, what would happen when an "inconsistant >route" >> >shows up? idealy, would that route be black holed? >> >Since it is "common" and since it "doesn't seem to break anything" in ral >> >terms, what happens? BGP advertises reachability to other BGP routers, be >> >they internal or external. But in terms of a packet traveling from my >house >> >to a destination that is "inconsistant" what happens? What matters? My >> >packet continues to be passed from here to there until some directly >> >connected router receives it. I'm assuming that "inconsistant" does not >> >imply "loop" >> > >> >thanks. >> >> You are correct in that inconsistent advertisements do not represent >looped >> routes. In the case of a prefix seemingly existing in two AS's, a remote >> router simply passes that prefix through the basic BGP path selection >> algorithm and selects the more preferable of the two for export to the >main >> routing table. Once a route hits the routing table, transiting packets >> are forwarded as usual. >> >> Any potential concern lies in the handling of routes that show up as >> inconsistent. I have seen discussions from various communities (RIPE >comes >> first to mind) about specifying a globally accepted behavior for such >> routes, but haven't seen a consensus on this issue other than to leave it >> alone. Howard probably has somewhat more detailed insight here. At >> present, inconsistent advertisements are accepted and many feel are valid >> and should not be handled differently from normal announcements. >> >> Customers who think that connecting to two providers is generally better >> than two pops from a single provider and providers who are too about >> nervous about losing customer revenue to force customers to properly > > multi-home (PI space/ASN) or not multi-home to different providers at all >> are likely the cause of this situation. So long as this continues to be >> the norm, we'll likely see more and more of these type announcements and >> the likelihood of routers dealing with them differently (dropping for >> example) will similarly decrease. >> >> Hit a route server (say route-server.exodus.net) and do a show ip bgp >incon >> and you'll see just how many of these routes we are dealing with. >> >> Pete >> >> >> >> >> >> > >snip for brevity< Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=45198&t=45198 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
