Illustrated TCP/IP by Matthew G. Naugle Wiley Computer Publishing, John Wiley & Sons, Inc. ISBN: 0471196568   Pub Date: 11/01/98

Chapter 104
Disadvantages of the RIPv1 Protocol

As noted earlier, the acceptance of RIP in the Internet community was based on its implementation into the popular Berkeley 4BSD Unix operating system through a process known as routed (Pronounced Route - d, two words). Unfortunately, it was implemented before the rapid growth of TCP/IP. RIP had many disadvantages that were not considered limiting at the time it became accepted. Before RIP was implemented, some router tables had to be constructed manually. RIP allowed these tables to be updated dynamically, which was a real advantage. The disadvantages follow.

RIP, based on a simple hop count, understands only the shortest route to a destination, which may not be the fastest route. RIP understands only hop counts. For example, there may be two paths to a destination: one that traverses two T1 lines (three hops) and another that has two hops, but is a 9600 baud serial line. RIP would pick the 9600 baud line, because its shorter (two hops). There are variations of RIP that allow the network administrator to assign an arbitrary RIP hop count or cost to a route to disallow for this. This solves one problem, but creates another. This incremented RIP number adds to the upper limit of a 15-hop diameter in RIP, which creates another problem. The number of hops that a network may be distanced from any network station is limited to 15—a hop count of 16 is considered unreachable. If you add additional hops to a path, you decrease the total number of routers allowed in a path.

With RIP, routing table updates are only as accurate as the router that submitted them. If any router made a computational error in updating its routing table, this error will be received and processed by all other routers.

What may also be apparent is the fact that the routing tables could get very large. If the network consisted of 300 different networks (not uncommon in larger corporations), each routing table of every router would have 300 entries. Since RIP works with UDP (connectionless transport-layer service), the maximum datagram size of a RIP packet is 512 bytes (576 bytes, including all media headers). This allows for a maximum of 25 <network number, distance> combinations in each packet. Therefore, it would take 13 packets from each router to broadcast its routing table to all other routers on all the local networks in the internet. This would be broadcast every 30 seconds by each of the 300 routers. All this, and the possibility that nothing had changed from the previous update! This is an unnecessary consumption of bandwidth, especially over slow-speed serial lines.

This leads to the second disadvantage. RIPv1 normally broadcasts (datalink physical address of all FFs) to the network every 30 seconds, even across slower-speed serial links. This makes the datalink pass the packet up to the upper-layer protocols on all stations on the network, even if the stations do not support RIP.

Every time we solved one problem, another popped up.