> We have two nodes, a "SEND node" and a "DIID node".
>
> SEND node:
> - (for simplicity) do not implement the DIID-style optimization
> - have an (e.g.,) EUI-64 based interface identifier, I.
> - configure a SEND (CGA) address P:A (where A is the identifier, A != I)
>
> DIID node
> - implement the DIID-style optimization
> - (for simplicity) do not implement SEND
> - happen to have A as an interface identifier
>
> The SEND node comes to a link. It perform DAD for both fe80::I and
> P:A, and confirms that these are unique.
>
No, the SEND node performs DAD for fe80::C, where C is a function of the key
and some other parameters. The prefix on LL addresses is also calculated
from the key.
> Then the DIID node comes to the same link. It performs DAD for
> fe80::A, and confirms it is unique. So the DIID node start using
> P:A without doing DAD while P:A is actually duplicate.
>
Yes, that's the concern. If the DIID node comes on the link second, then it
will assume that A is a unique prefix even though it only DADs it for the LL
address.
jak
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[EMAIL PROTECTED]
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
