Beruflich Dokumente
Kultur Dokumente
networks
A question has been raised on CCNA Study Group at CLN1 about RIP version 1 (RIPv1, RIP-1)
behavior. In my view, the point of that discussion was about autosummarization behavior in
contiguous and discontiguous networks. I think that was a good discussion and I felt motivated to
write this document, that might be helpful for anyone who may be curious about it. And I'd just say
curious because for new CCENT/CCNA exams (V2.0) RIP-1 won't be needed anymore, so the kind
of details discussed here might probably not be of interest to you.
All the captures here was made using Putty and GNS3 running Cisco IOS. This was the
topology I used:
As RIP-1 doesn't provide any field to carry the subnet mask information in its updates, the only
possible routes exchanged would be the classful ones (like 10.0.0.0/8, 172.16.1.1/16, 200.1.1.0/24,
for example) since RIP-1 is a classful routing protocol. This fact would be reinforced by RIP feature
called autosummarization, which cannot be disabled in RIP-1 (you can even type in the command
"no auto-summary" into CLI but it will have no effect at all).
Surprisingly, there are some cases in which RIP-running routers would receive and accept
routes for a subnet (like 172.30.115.0/24) instead of summarizing it to the major classful network
(172.30.0.0/16). Regardless of a subnet or a classful network, all that information - the routes – are
supposed to be added to router's routing tables and for this task there are some rules going on the
devices that help them decide between route to a subnet and route to a major classful network. The
question is: where is this decision made - at the sending router, by sending a route in one of those
forms, or at the receiving router, by applying a mask it thinks to be appropriate for a route's network
address received in a RIP update?
I'll try to show in the captures that the answer is "both things happen", that is, the sending router
already sends the route for the subnet (like 172.30.115.0, without the subnet mask information) and
the receiving router uses some logic to apply a proper subnet mask to that subnet. It's worth
mentioning here that the route might even be dropped by either or both the sending and receiving
routers (see the first two links below for this).
You can read about this subject in great books like Cisco Press' OCG, but for the details
discussed here I'll rely most on these (external) links from Cisco's website2:
- Behavior of RIP and IGRP When Sending and Receiving Updates
- Why Doesn't RIP or IGRP Support Discontiguous Networks?;
- Why Don't RIPv1 and IGRP Support Variable-Length Subnet Mask? (read about this at the
R3 has a loopback interface (L0) with the IP address 172.30.200.32/28. We will see this subnet
will not be advertised because of the rules of RIP-1.
We can confirm this by looking at R4's routing table:
result of this process can be seen in these two captures from R3 and R4 debugging:
Figure 5 - Debugging information on R4 showing it did not received subnet 172.30.200.32/28 from R3
Now, just to compare things, let's take a look at something slightly different on R3 that could
make it advertise a subnet 172.30.200.0 (we would change the subnet address for the sake of this
example).
If the subnet mask of L0 at R3 was changed from /28 into /24, then this subnet would be
processed as exactly the same as the subnet 172.30.115.0/24. The difference here would be that R4
would have a route to the subnet 172.30.200.0/24 instead of 172.30.200.32.
The next four pictures illustrate the result of this change, just for the sake of comparison.
5
Figure 8 - Debugging information on R4 showing an advertisement for the new subnet 172.30.200.0/24
Figure 9 - R4's new routing table with new subnet 172.30.200.0/24 advertised by R3
Another change that could make the subnet 172.30.200.32 be sent to R4 would be to change
interfaces in the subnet 172.30.100.0 for both R3 (S0/1) and R4 (S0/0), from /24 to /28, so that they
could have the same mask as 172.30.200.32/28. However, the implication of this change is that
subnet 172.30.115.0/24 would stop being advertised, since its mask would be different from the
R3's S0/0 subnet mask (/28).
Finally, this discussion also slightly draws attention to another topic since we are talking about
different subnet masks: VLSM. That's why I put the third link from Cisco's website, for as we all
know, RIP-1 has also the limitation of not supporting VLSM.