Re: [v6ops] DHCPv6/SLAAC Make Hosts Confusing-//RE: new draft: draft-liu-bonica-v6ops-dhcpv6-slaac-problem

[Mikael Abrahamsson <[e-mail]>]

On Thu, 24 Oct 2013, Victor Kuarsingh wrote:

> I understand Nick's point.  Coming freshly off a long sting at an ISP, I 
> see the value (if it can be done in a sound technical manner), of having 
> the * option * of a clean deployment for addressing and supplying IP 
> related configuration information to the end host/router.

When I first started working with IPv6 I didn't understand the separation. 
But then again, now that I have gotten used to it, I don't really see the 
reason to avoid RAs either.

On IPv4, the router person has to configure an IPv4 address and netmask, 
this then has to be mirrored by the DHCPv4 configuration.

On IPv6 you can do exactly the same, just set the A-bit off. As stated 
before, you even don't need an on-link prefix to be sent to clients, so 
this is actually *better* than IPv4, because if there is a 
forced-forwarding configuration on L2 the configuration can now be 
mirrored at the client so no "local-proxy-arp" is needed (which is an ugly 
hack).

So if one really wants to, one can exactly mirror the workflow of IPv4 
into IPv6.

If I was going to change anything, I wouldn't focus on RA/DHCPv6, but 
instead on ND. The whole "discovery" part is a mess, and it would be 
better if the router just could keep state with hosts what IPv6 addresses 
were in use and never sent out an discovery message. If it wasn't already 
in the Neighbor table then the packet would just be dropped.

-- 
Mikael Abrahamsson    email: [e-mail]