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

[Andrew Yourtchenko <[e-mail]>]

On Wed, 30 Oct 2013, Mark ZZZ Smith wrote:

> (Disclaimer: Haven't had a chance to read the text yet)
>
> I would generalise the description of RAs. 
>
> They're not "Router Advertisments", they're Advertisements (typically) 
>from Routers, and they propagate per-subnet IPv6 parameters to hosts - 
>per-link MTU, per-link ND timers, per-link address configuration method, 
>per-link on-link prefix status, default router lifetime (if not announced 
>as zero). The can be used to propagate per-client per-link IPv6 
>parameters if necessary using unicasts.
>
>
> I think DHCPv6 is really solving the application layer configuration 
>problem, which is independent of and not specific to the underlying links 
>and network layer parameters. I think IPv6 per-link parameter setting and 
>application layer parameter setting are conceptually quite different 
>problems (I think the sets of layer parameters they provide values for is 
>one of the give aways.)

I will try to rephrase this into another section:

-- cut --
Link-centric vs. client-centric

RA traditionally provide the information pertaining to the link and its 
parameters, whereas DHCPv6 typically provides the information specific to 
the client configuration.

However, this distinction is mostly a semantic one "by convention" - 
because technically nothing prevents to put a per-client configuration 
into an RA that would be sent unicast (at this point however the periodic 
multicast RA mechanism will break), as well as use the scopes in Stateless 
DHCPv6 with appropriate blobs of data to communicate the per-link 
parameter.

-- cut --

Is this an adequate representation ?

--a