Re: Comments on draft-wbeebee-ipv6-cpe-router-01.txt

["Ole Troan" <otroan@employees.org>]

> We are not even concerned about the draft being a WG work item.  We have
> not even completed requirements for the device.

perhaps you would have more success if you wrote down the requirements
instead of arguing for solutions.

>> no, you cannot use stateless DHCP for prefix delegation.
>
> Also, if you say IA_PD is forbidden by stateless DHCPv6, please point to
> a MUST NOT text in any RFC.

am I the only one thinking this discussion is becoming a little surreal?
_stateless_ DHCP is used in cases where  the DHCP server does not need
to keep and maintain dynamic state for the client. read RFC3736. PD
just as address assignment requires state.

>> e.g:
>> - false: downstream interfaces (loopback, LAN) can run SLAAC or DHCP
> IA/PD protocols
>
> The Loopback interface doesn't face the downstream or the upstream.  You
> and I both know RFC3633 has no explicit text that prohibits stateless
> DHCPv6 asking for IA_PD as I showed how (using the ORO once an interface
> has acquired a global IPv6 address).

I don't understand what you are trying to say here. you are alluding
to some model where the "loopback" (lets call this a virtual
interface) interface is bridged with the WAN interface. I have no idea
where you are going with that.
when I'm talking about a loopback interface or a virtual interface on
a router it is a L3 interface connected to a separate albeit virtual
link.
you cannot run PD on this interface for obvious reasons. it isn't
connected to a DHCP server.

>> - false: you need a global address on the WAN interface for uRPF
>
> If you don't see the obvious problem with uRPF with just a link-local
> address on the WAN interface, I suggest you go do a test and get back to
> us.

I don't see the obvious problem, nor do I see any other problems.
perhaps you can enlighten me?

>> - false: loopback interface is required. (it's only purpose is to have
> a stable always up management interface).
>
> DSL folks have asked to support the WAN interface acquiring a link-local
> address.  So it is the Loopback interface that acquires a global IPv6
> address to source packets to the WAN.  If the Loopback interface is not
> spawned, then since the LAN interface has global IPv6 address, one could
> say, use the LAN interface global address to source packets to the WAN.

how do you intend that a  (routed) loopback interface should acquire an address?
(as a matter of terminology, link-local addresses aren't acquired.
they are generated).

> But the CPE Router will not even function if the WAN interface has only
> link-local and the LAN interface has a global IPv6 address - see section
> 4 of
>
> http://www.ietf.org/internet-drafts/draft-miyakawa-1plus64s-00.txt

untrue, for the host function, routers in general implement the weak
host model. even the implementation alluded to in the draft can be
configured to do that. isn't the purpose of your draft to specify
_how_ a CPE router should be provisioned? or do you want to find the
lowest common denominator among existing IPv6 implementations...

/ot