RE: [dhcwg] Using DHCP to do link-local DAD?

Thomas, 

> From what I understand, netlmm is not chartered to extend/break the IP
> subnet model, and a number of people would have a lot of qualms about
> doing so. So I doubt seriously that netlmm is going to adopt multilink
> subnets.
> 
> Presumably you are aware of
> draft-thaler-intarea-multilink-subnet-issues-00.txt.

Right; I know about that one, and I am not pushing for multilink
subnets. I also know about Dave's draft on Multi-subnet MANETs
(draft-thaler-autoconf-multisubnet-manets-00.txt) and that is
a model the NETLMM could easily adopt without running afoul of
multilink subnet issues.

> I think it is extremely premature to be coming to DHC with this sort
> of problem, since I don't think netlmm should be going in such a
> direction, precisely because netlmm would be _creating_ problems that
> other protocols are then forced to deal with. This is not a DHC
> problem and shouldn't be made into a DHC problem.

I'm just asking about whether DHCP can be used to do DAD
for link-locals (i.e., instead of RFC2462 DAD). That
discussion doesn't need to bleed over into NETLMM,
multilink subnets, etc.

>> Well, some LLs like IEEE 802.11 don't have very reliable multicast,
>> and I don't think the IPoverFoo specs could fix that?
>
> Yes they could, if they thought that multicast was too broken to use
> for ND and other protocols. They could add a layer (e.g.,
> retransmissions) to increase the reliability of multicast.

But take IEEE 802.11 for instance. It has CSMA with collision
*avoidance*; not collision *detection*. So the sender doesn't
get any indication of multicast delivery failure and has no
way of knowing that a retransmission is needed.

Fred
fred.l.templin@boeing.com

Thomas

_______________________________________________
dhcwg mailing list
dhcwg@ietf.org
https://www1.ietf.org/mailman/listinfo/dhcwg