RA & DHCP problem...

[Lorenzo Colitti]

On Mon, Dec 30, 2013 at 5:07 AM, Brian E Carpenter <
brian.e.carpenter at gmail.com> wrote:

> What I'm seeing is Nick describing a use case in bits and pieces.
>  If we could have a draft that describes the use case completely,
> with example numerical requirements, that would be very helpful IMHO.
>

AIUI Nick has said that the main reason to he wants to put routing in
DHCPv6 is because his network will need to run DHCPv6 anyway and doesn't
want to use two protocols.

Not sure you can really call that a use case, since it's not something that
can't be done with RAs.

The failover considerations are not relevant since those depend only on the
router redundancy protocol (e.g., VRRP), not on the protocol used to
communicate routing information to hosts.

The fact that RAs cannot provision different information to different hosts
is also mostly a question of which protocol is used between the routers and
the configuration store (e.g., radius), not so much which protocol is used
between the routers and the hosts. For example, current access network gear
can already send different RAs to different clients by automatically
putting clients into different VLANs based on MAC address and fetching the
VLAN information from radius.

So - again AIUI - the argument is mostly the usual "DHCPv6 is more suited
to this network than RAs". This is true, but there are other networks where
RAs are a better fit than DHCPv6. The question has always been whether we
should completely duplicate routing configuration functionality into
DHCPv6, and add support for that into clients, and the answer has
repeatedly been that there is no consensus to do so.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cluenet.de/pipermail/ipv6-ops/attachments/20131230/35862e2c/attachment.htm>