Re: [dhcwg] DHCPv6 router option

[Iljitsch van Beijnum <[e-mail]>]

On 11 mrt 2009, at 18:06, [e-mail] wrote:

>> That's why I suggested having the DHCPv6 server tell the client WHICH
>> of the available routers to select, but if and only if that router
>> announces its presence.

>> This way, pointing towards a black hole and hosts losing connectivity
>> can't happen.

> I am considerably more worried about a possible inconsistency between
> the router and the DHCP server, than pointing the client towards a
> black hole. It is not a problem for us today, and I see no reason why
> it should be a bigger problem with IPv6.

There are problems with DHCP in IPv4 all the time, especially for non- 
expert users. Having routers announce their own presence was one of  
the things that make IPv6 more reliable than IPv4. Throwing away that  
feature is completely unacceptable to me, so expect unrelenting  
opposition if that is what you intend to standardize here.

However, the underlying need here is for administrators to make sure  
that the hosts send their traffic to the right router, and don't just  
randomly select any address for which there are RAs. This need is met  
by allowing the DHCPv6 server, if present, to impose the selection. So  
I don't see how this change takes away anything from the intended use.  
This just requires setting up the designated router to send out RAs  
(easy), without requiring making all other routers stop sending RAs  
(hard).

Yes, this means that you still need to have RAs in your network. In an  
ideal world you would be able to get rid of those if you don't want  
them. But this isn't an ideal world, and RAs don't get in the way in  
any way that I can see, so having to live with this is more than worth  
having the additional robustness that hosts can still talk to the rest  
of the network if the DHCP server dies or someone fatfingers its config.