Re: Enabling SLAAC for IPv6 by default

[Roy Marples <roy%marples.name@localhost>]

On 25/09/2018 23:49, John Nemeth wrote:
> On Sep 25, 10:41pm, Joerg Sonnenberger wrote:
> } On Tue, Sep 25, 2018 at 03:56:09PM -0400, Greg Troxel wrote:
> } > So perhaps the kernel code should be modified to support, and prefer by
> } > default over the original SLAAC, the RFC7217 mechanism, to address the
> } > concern Pierre raises for the case when someone uses the in-kernel
> } > SLAAC client?   Or perhaps the situation we have, where it is disabled
> } > by default, is ok?
> }
> } I'm personable more in favor of removing the in-kernel RA support. I do
>
>       I am vehemently opposed to the idea of removing in-kernal RA
> handling.  RA handling is fundamental to the IPv6 protocol.  You
> should not have to run a userland daemon in order to be able to
> use IPv6.  My IPv6 boxes are perfectly fine without any daemons
> for basic networking (other then rtadvd for the router).  I would
> not be happy to see this situation change.

In a pure IPv6 environment how do you populate resolv.conf from the DNS 
options found within the RA without a userland daemon of some kind?

>       Like certain others, I'm still annoyed by the removal of
> rtsol{,d} as it was a simple program for doing a single very simple
> task.  However, in the end, it just means that IPv6 takes a bit
> longer to come up.

Who are these others? Generally it's only you I hear moan about this.

So simple it lacks some important functionality:

How does rtsol ask dhcp6 to start based on the O/M flags?

How can you tell the kernel to prefer the medium preference router on 
the wireless interface to the high preference router on a cable interface?

How can .... I could go on, but you get the idea.

Just because a simple program and the simple in kernel RA handling works 
fine for you does not mean it works fine for others.

Roy