[mif] RA vs DHCPv6 config (was Review requested: draft-ietf-mif-dhcpv6-route-option)

[Brian E Carpenter <[e-mail]>]

Ted,

On 2011-10-31 06:24, Ted Lemon wrote:
> On Oct 30, 2011, at 1:02 PM, Alexandru Petrescu wrote:
> The question one raised on 6man is about coexistence with RA about
> default route.  One is aware that a similar situation (alternate
> mechanism DHCP-vs-RA for default route) appeared recently when DNS-in-RA
> was proposed (DHCP existed doing DNS).  RFC6106 proposes to  do
> DNS-in-RA but has a section explaining coexistence with DHCP about DNS
> address - and gives the latter precedence over.
> 
> This is a very good point, which should be addressed in the route option draft.   I think the right thing is to give RA precedence over DHCP for routing information, but am curious to know if others disagree.

It makes my head hurt a bit to give RA config priority in one context and
DHCPv6 config priority in another context. I think this point requires
a wider discussion; you could find opinions about this in (at least)
6man, v6ops, homenet, and 6renum.

> In some cases this recommendation may be inappropriate - there may exist
> cases where routing protocol software _and_ DHCP software should be used
> on the same machine (e.g. use DHCP to get DNS address, and use OSPF to
> do routing).  At that point it may be hard to prevent some particular
> option of DHCP (route-option) being physically available on the machine.
> Accidentally misconfiguration may happen.
> 
> Fortunately, this is not a very serious problem: the router and the DHCP server are both under control of the administrator, so they can simply configure them correctly, and the right thing will happen.   It is always possible, if the network administrator sets things up wrong, for the network to not work, and there is nothing the IETF can do to eliminate this risk.   Since the default case is for the network administrator not to configure DHCP, I think it's pretty safe to assume that we won't get a bad route configuration without some kind of positive action on the part of the administrator.

Not necessarily, if you think about out-of-the box default behaviour
when someone hooks a few boxes together in an unmanaged network. That's why
I have a feeling that "in case of conflict, DHVPv6 always wins" may turn out
to be the right answer.

    Brian