Alain Durand wrote:
Coming late into this discussion...
In our deployment which concern a very large number of devices (many
millions), we will use DHCPv6 only.
Thanks for your answer, it certainly adds much to the poll results. (!!)
Would you mind a few follow-up questions, on the DHCPv6 set-up you
described? I'm sure I'm not alone in being interested in the details...
Will each CPE device's "clients", i.e. customer hosts, be directly
assigned addresses, with the CPE acting as a relay? And thus hosts in
one head-end receive assignments from a prefix on a first-come, first
serve basis?
Or, will the CPE be delegated prefix(es), and hosts given addresses out
of the appropriate delegated prefix?
What prefix length(s) will be in use, typically?
How are addresses assigned and/or constructed? Does client MAC get used
directly (EUI-48) or indirectly (EUI-64) or not at all?
Is there anything related to DHCPv6 that your set-up either supposes, or
supports, related to prefix delegation to customer-operated CPE devices,
like stateful firewalls?
(I'm thinking (cloud)->CPE docsis3.0 modem -> stateful firewall -> (N *
customer devices).)
BTW - I'm not trying to question your network design or implementation,
or anything remotely similar.
I'm only trying to understand whether anything from DHCPv6 is in any way
limited in what kind of prefixes it can use, or whether anything is
steering you towards specific prefix length(s).
Any light you can shed, as far as implications on prefix-length vs
DHCPv6 is concerned, would be appreciated.
The reason I ask is, I'm working on stuff to address prefix-length
issues related to autoconfiguration.
Given the number of clients you are talking about, I think I need to be
sure any relevant DHCPv6 stuff gets handled too.
Thanks,
Brian
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------