[ipv6hackers] DAD and Link-Local

[The Fungi]

On 2012-07-13 14:49:37 +0100 (+0100), daniel.bartram at bt.com wrote:
[...]
> Aware that DAD is performed on link-local as well as global
> addresses to ensure uniqueness

Well, more specifically DAD is required of any SLAAC implementation
if it's going to be RFC-4862 compliant. If you're assigning your
addresses via some other mechanism (including manually) then DAD
very well may not occur.

> but if the link-local is a randomly generated value, or even the
> IID for that matter, on a P2P link using only link-local
> addresses, what are the chances of those addresses ever matching?

For EUI-64 addresses, the two should only ever match if the devices
had the same MACs. How likely that is depends mostly on
unpredictable human factors (posibilities of MAC reuse within a
given IAB/OUI or custom assignment of interface MACs). For a random
address generation algorithm operating across a /64 with perfectly
even distribution, the chances of a collision between two addresses
without DAD would be 2^64 (something like a one in 18 quintillion
chance).

> In which case, can DAD be disabled?

If the address auto-generation implementation allows DAD to be
disabled (for example, RFC-4862 section 5.4 describes this), it's
entirely at the option of the system administrator to do so.
-- 
{ IRL(Jeremy_Stanley); WWW(http://fungi.yuggoth.org/); PGP(43495829);
WHOIS(STANL3-ARIN); SMTP(fungi at yuggoth.org); FINGER(fungi at yuggoth.org);
MUD(kinrui at katarsis.mudpy.org:6669); IRC(fungi at irc.yuggoth.org#ccl); }