Network Addresses

IPX (Internetwork Packet Exchange) is developed by Novell based on work by Xerox.  The network address is 80-bit, with 32 bits for network part and 48 bits for host part.  The host part is simply the MAC address.  The network address is constructed by concatenating the network address obtained from the router periodic boradcast and the MAC address.  When sending a packet, IPX will check the network part if the packet is for the same subnet.  If it is not, the packet will forward to the router.

DECnet Phase IV addresses are 16-bit long.  6 bits are for the network part and 10 bits for the host part.  The MAC address is modified by replacing the last 16-bit with the DECnet address.

Appletalk uses 24-bit addresses. 16 bits are used for network part and 8 bits for the host part.  Similar to IPX, the network part is learnt from the router.  For the host part, the host will pocks an address and checks if it is being used by sending out a message to this address.  If there is no answer, the host will use this address.  If the address is being used, the host picks another address and retries.  Appletalk use an address resolution protocol similar to ARP.

OSI (Open System Interconnection) protocol family was created by ISO and ITU in 1980.  The Connectionless Network Protocol (CLNP) provides a datagram service (CLNS).  X.25 is the connection-oriented part of the protocol family.  CLNP addresses varies in length with maximum length of 160 bits.  The MAC address is often used as part to CLNP address.  CLNP host always sends packet to routers and not trying to figure out if the target host is in the same network or not as would other protocols do.

TCP/IP was created by researchers connected to the US Department pf Defense Advanced Research Projects Agency.  Originally, ARPANET used a single Network Control Protocol (NCP).  NCP was splited into IP and TCP at round 1980.  IP was intended to connect network of different types and thus it uses a fairly short address scheme of 32 bits.  IP addresses were divided into 3 classes to accomodate different network configuration.  Unfortunately, class B addresses was so popular and used up in 1990s.  Multiple class C addresses were given out to temporarily fixing the class B shortage.  This problem was fixed by the introducing the Classless Interdomain Routing (CIDR) in 1993.  A value is used to define the exact split of network and host part is carried in the routing.  Becasue there is no relationship between the IP address and the MAC address, a protocol (ARP) is used to identify the binding.

IPv6 uses 128-bit addresses and is fully classless.  In most cases, the first 64 bits are used for network part and the remainign 64 bits used for the host portion.  The host part by default contains the EUI-64 (Extended Unique Identifier) which is simply the MAC address.  This 64-bit MAC address contains the usual 48-bit MAC address with extra 15 bits of 1s and 1 bit of zero inserted in the middle.  The network part is derived from the regular router broadcast.  IPv6 also supports manually configured address or DHCP.  Because MAC address may not present in the IPv6 address, IPv6 uses ND (Neigbhour Discovery) protocal to identify the host on the same subnet.  Unlike ARP, ND uses multicast instead of broadcast