In IPv6, you can use different types of addresses. The following is a summary of these address types:
Link-local addresses: These addresses can be compared to APIPA addresses in IPv4. These addresses are used if no specific information about the network configuration could be found. You can consider them IP addresses for local use only. These addresses always start with FE80 in the first 2 bytes. They are not routable, but they are necessary for neighbor discovery (see later in this section). Link-local addresses are always created automatically if IPv6 is enabled.
Site-local addresses: You can compare these addresses with the addresses that are defined in the private address ranges for IPv4. Site-local addresses always start with FEC0 and have a default 48-bit subnet mask. You can use the last 16 bits for internal subnetting. Site-local addresses are not created automatically.
Aggregatable global unicast addresses: These are the "normal" addresses that are used on IPv6 networks. They are assigned by an administrator and always start with 2 or 3 (binary 001).
Multicast addresses: These are addresses used to address groups of nodes. They always start with FF.
Anycast addresses: These are the IPv6 alternative for a broadcast address. When using anycast, the IPv6 node gets an answer from any node that matches the anycast criterion.
In IPv6, broadcast addresses are not used.
On a single Linux host, you will always find more than one IPv6 address:
• Aloopback address (::1) is used on the loopback interface.
• A link-local address is generated automatically for every interface.
• If the administrator has configured it, every interface has a unicast address. This can be a site-local address, an aggregatable global unicast address, or both.
Was this article helpful?