I expect this comment relates more to IBGP where direct reachability is not the norm.
At 06:55 PM 4/6/2002 -0500, Priscilla Oppenheimer wrote: >At 01:10 PM 4/6/02, JohnZ wrote: > >"IP connectivity has to be achieved via a protocol different from BGP; > >otherwise, the session will be in a race condition. An example of a race > >condition follows: neighbors can reach one another via some IGP, the BGP > >session gets established, and the BGP updates get exchanged. > >This must be referring to external BGP peers that are not directly >connected and must rely on an IGP for connectivity? (EBGP multihop?) >Normally BGP peers are on a LAN or WAN and can reach each other directly, >simply with IP and TCP. I know that's not your question, but I thought it >was worth mentioning that the example is kind of a weird situation. > > > The IGP > >connection goes away for some reason, but still the BGP TCP session is up > >because neighbors can still reach each other via BGP. Eventually the session > >will go down because the BGP session cannot depend on BGP itself for > >neighbor reachability" > > > >Wouldn't the same condition occur if reachability is acheived via a > >different protocol. > >If the other protocol fails, yes. That's what he's saying. The race >condition occurs when BGP must rely on BGP for connectivity. This should >never happen. The routers can normally reach each other directly with >TCP/IP or via an IGP. But if the BGP protocol had been written such that >BGP depended on BGP, that's a race condition. It's also a race condition >when the other protocol fails and BGP is once again depending on BGP. > >I hope that makes sense. > >Priscilla > > >If the route becomes unreachable then BGP conectivity > >will still be lost.What's the advantage of making sure that "race condition" > >is avoided. > > > > > > > >Thanks. > > > >JZ >________________________ > >Priscilla Oppenheimer >http://www.priscilla.com Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=40721&t=40690 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
