Classless Inter-Domain Routing (CIDR)

IP addresses have been running out so moves have been made to find a solution to this. One solution is to use private addresses in conjunction with NAT (see later), another long term solution is to develop a new version of IP, IPv6 that uses 128 bits for the address rather than the current rather limiting 32 bits. A third solution is CIDR. CIDR was first mooted in 1993 in RFC 1517 and modified in RFC 1518, RFC 1519 before being finalised in RFC 1520.

CIDR has been designed to allow us to make the best use of the IPv4 addresses that exist. If you have a requirement for 2000 IP addresses then you would not have enough if you were just given a Class C address (which gives 2⁸ - 2 = 254 hosts). If you were given a Class B address you would then have enough addresses for 2¹⁶ - 2 = 65534 hosts, however you are never going to need all these and because you are the owner of these addresses no one else can use them. This makes for inefficient use of the IPv4 address space which is rapidly running out of room.

Nowadays there are very few free Class B addresses, but there are some Class C addresses. It is common for ISPs to be issued Class C addresses in blocks so that they can issue them in contiguous address blocks to companies that require more than 254 IP addresses. For the above example, if you were given 8 Class C addresses such as 199.103.104.0, 199.103.105.0 .... up to 199.103.111.0, then you would have 8 x 254 = 2032 IP addresses. If you look at the third octet you will see the following:

In bold, the bits do not change, you could therefore group these 8 Class C addresses together as the network address 199.103.104.0/21, since 21 bits form the network mask that encompasses the 8 Class C addresses. Normally, when you see the 199 in the first octet, you expect to see this as one Class C address with a 24 bit mask. What has happened in effect is a summarisation, but instead of a proper summarisation in a variable subnetted environment where the summary mask sits somewhere between the natural mask and the subnet mask, this summarisation has occurred beyond the 'class boundary' and is called Supernetting.

Officially, Class C address ranges such as 192 and 193 have been given to ISPs in the US as 'Classless'. Although these were originally Class C addresses, they can now be treated as Class A address ranges and organisations can now be issued with more flexible amounts of addresses rather than be restricted to the class boundaries. Europe has been assigned the address ranges 194 to 204.

The following table details the CIDR address allocation by region:

By regionalising these Class C addresses into blocks different sized CIDR blocks can be summarised up the Internet hierarchy. So blocks with prefixes /23, /22 etc. can be summarised into larger blocks such as /19, /18 etc. In fact is common for NAPs to not accept blocks that are less than /19 in size in order to minimise administration and maximise route stability.

In the above diagram, address ranges A, B and C are in sequence. ISP 2 has to be careful when summarising, not to include address range A with address range C, they must be kept separate. The problem is that packets bound for address range B would think that they need to go to ISP 2 when in fact ISP 1 owns that range and the only route to B is via ISP 1. This is known as a Black Hole.

See RFC 2050 for more information on address assignment.

Classful routing protocols such as RIPv1, EGP and IGRP will default to the natural mask (or classful mask) for a network in it's routing table or will use the interface configured mask (this requires the mask to be the same throughout the IGP AS however!).

More information on CIDR can be obtained from the www.internic.net website.

Home