Poisoned Reverse - Loops of size greater than 2

Question

It is said that poisoned reverse can prevent routing loops, but only those of size 2. Why is it that it cannot prevent routing loops of a larger size? In other words, is it still possible for looping to occur even with poisoned reverse? I've tried looking it up on the net, but I have yet to find an example.

Answer 1

A ¯¯\
|    C———D
B __/

In the figure above(I am not able even to upload an image with my 3 point reputation now). Now C-D fails, and suppose the original optimal path from B to D is B-A-C-D, which means B will advertise C this optimal path from B's view.

In this case, even with poison reverse, C can pick B as next hop for D. A loop is formed again.

I also tried to search an example on the net... so I found your post :), but I finally find an example at the following link:

http://www.mpi-sws.org/~gummadi/teaching/sp07/datanets/homework/homework2solution.pdf