These terms aren't defined authoritatively anywhere, but thinking about the English-language meaning of the terms does help. Poisoning a route simply means stating that a route is unreachable. If it's sent in the reverse direction, then it's poison reverse.
Poison reverse is usually used as a form of split horizon in a proactive manner, as I described for EIGRP. But the term doesn't have to be used that way. Cisco claims to support poison reverse for RIP. But the behavior is as I described and not proactive (at least on routers; I haven't studied routing switches). For any definition that you find that describes precise behavior, I can find another one that describes it differently! ;-) The terms are used in a variety of ways. Priscilla At 01:25 PM 1/28/02, s vermill wrote: >Priscilla, > >Isn't there a difference between poison reverse (which is a variation on >split horizon) and route poisoning? I thought poison reverse took place >each time a route was learned. For example, router A advertises network 1 >to router B. Router B immediately poison reverses the route to router A. >Route poisoning only happens in the triggered fashion that you described in >your post. Or so I thought. > >Thanks, > >Scott > >Priscilla Oppenheimer wrote: > > > > Cisco does actually support a form of poison reverse with RIPv1 > > and v2. > > It's not proactive, but it still fits the definition. > > > > When a router loses a route to a network behind it, it > > announces that it > > can't get to the network by sending a RIP update that lists the > > network > > with a distance of 16. The router does this quickly, without > > waiting for > > the next update timer. It sends a triggered update. > > > > I have observed that other routers then also say, "hey don't > > use me to get > > there either." These are routers that never could have gotten > > there anyway > > without the help of the other router. Their only path was in > > the reverse > > direction from the networks that they advertise. > > > > An example might help: > > > > ---network1----Router A-----network 2------Router B-----network > > 3--- > > > > Router A loses its route to network 1. Router A sends a > > triggered update > > listing network 1 as unreachable (hop count = 16). Router B > > then also sends > > an update listing network 1 as unreachable (hop count = 16). > > That could be > > considered a poison reverse feature. > > > > Note that this isn't as proactive as some forms of poison > > reverse. If this > > were EIGRP, for example, as soon as Router A announced it could > > get to > > network 1, Router B would send an update saying its distance to > > network 1 > > is infinity (delay = max). It proactively tells Router A not to > > ever use > > Router B to get to network 1. Router B says it is not a > > feasible successor > > for that network. That's definitely poison reverse. Cool, eh? > > > > Bottom line: Cisco's implementation of RIP (and of course, > > EIGRP) has > > always been a bit more advanced than the textbook descriptions > > of a > > distance-vector algorithm. > > > > Priscilla > > > > At 11:07 PM 1/27/02, Pierre-Alex GUANEL wrote: > > >Cisco does not seem to support poison reverse for RIP and RIP > > version 2. > > > > > >Do you know network vendors who do? > > > > > >Pierre-Alex > > ________________________ > > > > Priscilla Oppenheimer > > http://www.priscilla.com ________________________ Priscilla Oppenheimer http://www.priscilla.com Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=33478&t=33402 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
