As a distance vector routing protocol, we do not want to see whether it is a routing loop or slow convergence. Of course, it also has some auxiliary measures to remedy, such as: 1. Split-horizon (horizontal split) 2. Split-horizon with lateral ONED reverse (horizontal split with toxicity reversal) 3. triggered Update (trigger update) 4. holddown timer5 and define a maximum hop count (maximum value defined) however, the NA textbook does not clearly explain horizontal segmentation and horizontal segmentation with toxicity reversal.
1. Horizontal Split) Horizontal segmentation refers to the route information received from one end and cannot be sent back from the original path. As shown in: ---------- R1------------R2------------R3-----------10.1.1.1 172.16.0.0 192.168.0.0 we take a unified rip example: If R2 learned from R3 about the network segment 192.168.0.0 route update, can not send back to R3. We can use the debug IP rip to observe route updates on R3, and we will find that R3 sends the entries of the network segment 192.168.0.0 to R2 in the form of broadcast, however, the route update received from R2 does not contain entries in the CIDR Block 192.168.0.0. In fact, this entry is blocked or filtered.
Ii. Horizontal Split with toxicity reversal) However, it is always much better to receive even bad messages than to receive messages. We can achieve this using horizontal segmentation with toxicity reversal. In this case, open Debug on R3 and observe the rip update process. In fact, it can receive updates about R2, but the UPDATE contains 16 hops for the network segment 192.168.0.0, Which is infinite! So in any case, R3 will not place this entry in its own route table. This ensures the reliability of routing information. We recommend that you use horizontal segmentation with toxicity reversal.