[ipv6-wg] [Snac] [v6ops] "router cascade with DHCPv6-PD"

Hi Martin,

Thanks for the feedback. I've included your questions below.

> LPD-4: The IPv6 CE Router MUST give out a IA_PD of the remaining prefixes
with a IA_PD size of 64.

Why not honor the IA_PD size requested by the requesting router? Do you
expect that possible another cascade router would be switched to DHCPv6
relay mode?
[TW] - Yes. We expect Routers getting a bigger then /64 to be DHCPv6
Server, typically at the edge to ISP.  We are suggesting other routers
getting IA_PDs of size 64 on the WAN interface to act as relays on the LAN
interface.

> LPD-6: If a IPv6 CE Router recieves IPv6 prefixes of size 64 it MUST
enable delegating relay according to [RFC8987]. The MUST forward all the
DHCPv6 Relay messages to the All_DHCP_Relay_Agents_and_Servers multicast
address.

Manually configured unicast address of upstream DHCPv6 server would not
also be acceptable?
[TW] - RFC 7084 had the goal of zero configuration requirements for CE
Router.   I was trying to keep to that by suggesting using the multicast
address as the requirement.   We could make it SHOULD if we feel that's
more appropriate.

I'll look into what requirements from the DHCP Relay draft we should call
out.

~Tim

On Tue, Jan 17, 2023 at 2:48 PM Michael Richardson <mcr+ietf at sandelman.ca>
wrote:

>
> (Hi Martin, I'm just the messenger, your questions are all pretty good.
> For
> others, I think it was not obvious that they are about
> draft-winters-v6ops-cpe-lan-pd-00.
> But, more than my brain can handle right now)
>
> Ted Lemon <mellon at fugue.com> wrote:
>     >> Why not honor the IA_PD size requested by the requesting router? Do
> you
>     >> expect that possible another cascade router would be switched to
> DHCPv6
>     >> relay mode?
>
>     > What’s the use case for this behavior? And yes, we expect routers to
> act as
>     > DHCPv6 relays.
>
> I think that actually we've created an ambiguity about whether they are
> DHCPv6 relays or if they are full DHCPv6 servers.
> There are arguments both ways, and in some cases the choice is transparent
> to
> the end system, but in other ways it is not.
>
> The lack of a standardized way to communicate DHCPv6-server PD to
> DHCPv6-relay, and the expectation that the relay would just *snoop* on the
> DHCPv6-PD contents has bugged me for a long time.  If we are going to
> recommend that they act as relays, then I'd like to fix that.
>
> Either way, Tim's v6ops draft-winters-v6ops-cpe-lan-pd-00 should clarify.
> LPD-1 suggests the IPv6 CE Router is a server, not a relay.
>
> --
> Michael Richardson <mcr+IETF at sandelman.ca>   . o O ( IPv6 IøT consulting )
>            Sandelman Software Works Inc, Ottawa and Worldwide
> --
> Snac mailing list
> Snac at ietf.org
> https://www.ietf.org/mailman/listinfo/snac
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: </ripe/mail/archives/ipv6-wg/attachments/20230117/7fa3ee0c/attachment.html>