IPv6 Course part 3: The target- and Source IP Selection (prefix policy)

In IPv6, it is provided in the standard, that a computer, even closer 1 NIC, has multiple IPs. This may be due:

1. Each IPv6 network card has an IPv6, which is not globally accessible from the Internet, but only from physically connected peers (“scope: Link local”) _.
2. If IPv4 and IPv6 are enabled, then there are already 3 IPs (IPv6 Link Local, IPv6 global, IPv4).
3. The router can announce multiple prefixes (Subnets) _. The computer then generates an IPv6 for each of the announced subnets.
4. If the “privacy”-Extensions are enabled to anonymize, then it can be much more.
5. You can also manually set additional IPv6 addresses.
6. A computer can be reached via several different routes to the Internet or reach other subnets.

Two tasks resulting:

• If the DNS server offers multiple IPs for a computer,, then one must be selected it as destination IP.
• A source IP must be determined from the own IPs.

How to solve these tasks, is described in RFC 3484 http://www.ietf.org/rfc/rfc3484.txt _; the key points I presend here and explain the procedure with examples in a subsequent part of the course.

First, the target IPs are sorted so, that the probably best destination IP is the first. An application should then try, to connect to the destination IP. For this, then typically the operating system determines the source IP. If the connection to the best destination IP does not work, the application should try the other target IPs in the given order (this is often programed the other way around: First select the source IP for all possible target IPs and then choose the best destination IP. The reason for this is, that to choose the best destination IP eventually requiers to know. _, which source IP is used for this destination IP, see rule No. 3 and 5).

The sorting of the target IPs is based on following rules:

1. Avoid destination IPs, that are known to the computer, to be unreachable (p.E.. because the packet must be sent on a network card , where the cable is not connected at the moment )
2. If there is a source-IP with the same scope (“scope”) _, then take this destination IP.
   Multiple scopes are defined in IPv6:
   (in ascending order from local to global)
   

   01: interface-local
   02: link-local (FE80::1::/10)
   03: subnet-local
   04: Admin-local
   05: site-local (You start with FEC0, obsolete with RFC 3879. The new FD00 addresses, replace the 192.168.x.y, get the scope “global”)
   08: Organization-local
   14: global

   Since the FD00 addresses as “global” be classified, come mostly “02: Link-local” (the FE80 addresses, the mainly used in the auto-configuration) and “14: global” before. You can see the scope of the current IPs with

   # Windows
   if IPv6

   # Linux
   ifconfig

   This rule is practically, that

   • If one of the target IPs with FE80 to FEbf (= FE80::1::/10) begins, then she use.
   • Otherwise use a destination IP with scope “global”, If the computer is even a “global” (scope) Source IP has. If it's like in our case - multiple target IPs with scope “global” _, will be decided according to the following rules.

3. Avoid outdated source IPs (What “obsolete IPv6 addresses” exactly are, is explained in a later section on anonymous IPv6 addresses).

4. Rule for mobile IPv6 (This one keeps the IPv6 Home, Although it is on the way) -because this is so far very rarely used, not I explain it here.

5. Prefer the destination IP, in the label of the source IP with the label that matches the destination IP.
   At this point, the choice of the destination IP through a change to the address selection table - for example, as in part can 2 described - to be changed.

6. Prefer the destination IP, has a higher priority
   At this point, the choice of the destination IP can be changed by a change to the address selection table.

   Only if now is more than 1 IP available, the following rules are applied.

7. Prefer a native transport route (If the target either by a tunnel, p.E.. by Sixxs, to achieve or directly, p.E.. by IPv4, then use the direct connection)

8. Choose the destination IP, which “most lokalsten” is, meaning. the smallest value in the scope table.

9. Otherwise: Select the destination IP, where has the prefix of the most consistent with the source IP, meaning. left the most bits are the same.

These rules are used in the next part, to understand the standard table for address selection.