Re: To DAD or not to DAD?

>
>Hesham,
>
>Might make sense to read draft-yourtchenko-6man-dad-issues. Of the 11+
>issues/sections I think there are only two that are about efficiency.

=> That’s right, I did read it before opening my mouth :). I only
commented on efficiencies because it stuck in my head after looking at
your slides. I think I made more comments on other issues in another email
to Ole.  
I see those issues falling into 4 categories:
1. Increasing the scope of DAD’s support (e.g. supporting
partitions/joins, multi-link subnets ..etc), which were known “out of
scope” features before.
2. Problems which are not true because we have documented solutions for
them (e.g. delays before using an address)
3. Strange link layer behaviour. There is only one issue that struck me
there related to link layers that appear to be up but are not really up
because they’re running STP (issue 3), which strikes me as a misleading
“link up” indication, which can cause more problems than just DAD.
4. Implementation bugs (issue 5 only)

I guess as always YMMV on whether a valid problem is worth solving (i.e.
category 1 above) and the group decides that based on experiences. I only
commented on the ones that I didn’t agree with as problems.

DISCLAIMER: Ole sent a few links in addition to the draft, which I haven’t
read yet. 

>
>For instance, if we depend on DAD it would make sense for it to have
>better tolerance against partions/join on the link.

=> Right, that falls under category 1. For me that depends on data showing
how serious a problem this is. Personally I don’t know.

>
>One of the potential solutions might to treat large wireless systems
>(WiFi ;-) as point-to-point links.
>But before getting into solutions I think it would be good for the WG to
>look over the set of issues with DAD.

=> Agreed. We need to be on the same wavelength when it comes to the
actual problems. 

Hesham

>
>    Erik
>
>>
>> So IMHO yes collisions should be rare but I just don’t see DAD as this
>> major cost given the tools we have in various scenarios. So go to the
>> beach! It’s warm here so the beach is fine in the southern hemisphere :)
>>
>> Hesham
>>> --------------------------------------------------------------------
>>> IETF IPv6 working group mailing list
>>> ipv6@ietf.org
>>> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
>>> --------------------------------------------------------------------
>>
>>
>