> > Proposed resolution: write "MAY be disabled" instead.
> 
> Um, I'd suggest that this by itself isn't really all that helpful. It
> might help to add some more explanatary text together with
> recommendations on what to do. E.g.,
> 
>  - should an implementation just configure the address and use it
>    anyway? (I doubt this, as this implies we should just punt on DAD)

No, we clarified later that the implementation should disable the
address.
 
>  - Assume there is a DOS attack, wait a while, and try again? (hoping
>    the bad guy goes away in the meantime)

I would not specify that..

>  - generate a new IID and try again? And if so, are there
>    recommendations on what kind of ID? And if there is a  DOS attack
>    in progress, why would  trying *any* address result in success?

Yes. The implementation may very well have access to several EUI-64, and
may try an alternative. Or, it may try to use a random address. The
persistence of DAD collision would indicate with quasi certainty a DOS
attack. Forcing the attacker to repeat the attack increases the chances
of detection.

>  - Try generating a new address, but use exponential backoff in
>    delaying after each failed attempt?

That is a variation of the above. It may be useful in some cases, e.g.
when turning of the interface briefly saves energy.

>  - something else? (or some combination of the above?)

Maybe. There is no limit to creativity.

-- Christian Huitema

--------------------------------------------------------------------
IETF IPv6 working group mailing list
[EMAIL PROTECTED]
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to