[dhcwg] Re: dhcpv4-over-dhcpv6-ra: source addrs / server ID

[David 'equinox' Lamparter <[e-mail]>]

On Thu, Jul 25, 2024 at 07:50:09AM -0400, Bernie Volz wrote:
> router (likely also relay) would need an IPv4 address anyway.

The draft suggests implementing 4o6RA on an L2 switch.  With option 82,
that switch was previously inserting it on passing through an IPv4
packet - it didn't need an IPv4 address itself, and it likely doesn't
have one, at least in the "user" network.  (It might have one in an
isolated management network, but that might be IPv6, so the switch has
no IPv4 at all.)

Now the L2 switch translates incoming IPv6 4o6 replies into IPv4 for the
legacy client.  The 4o6 packet has no IPv4 source address;  the L2
switch needs to come up with one.  Is that going to be 0.0.0.0?

> And, we have existing mechanisms such as server-id override (added to
> v4 dhcp request in option 82) that could be used should the clients
> not be able to directly communicate with the dhcp server itself.

The draft might want to include a recommendation to use 5107;  if I
understand correctly it'd mean the DHCP server doesn't need an IPv4
address at all?  (Even if it still has one, the client bypassing the
relay and RENEWing with the server is probably not desirable,
considering this draft is all about topology information.)

(Also the multiplication of above 2 things clashes again, how do you do
5107 with an L2 switch that has no IPv4 address itself?)

Cheers,

equi
(David)