Re: [v6ops] [dhcwg] IPv6-Only Preferred DHCPv4 option

[Lorenzo Colitti <[e-mail]>]

On Thu, Dec 5, 2019 at 4:55 PM <[e-mail]> wrote:

> > I agree that in theory pref64 presence is a very good implicit signal
>
> for 'you don't really need IPv4'. However when a host connects to a
> > network it does not know if pref64 is available when it starts DHCPv4
> > process. So to use RAs as a signal you need to either delay DHCPv4 (==
> > penalize v4-only and dual-stack networks) or do all those undesirable
> > RELEASE tricks later.
>
> [Med] The machinery will be almost the same as you are doing with your new
> option. You don't need to delay DHCPv4. The only difference is the signal
> on which to rely to make the decision.
>

I don't think that works. What is needed is a signal to the host that IPv4
should not be used if the host is capable of IPv6-only operation. That
signal *must* be available before DHCPv4 assigns an address to the
interface, because once an address is assigned, the host will not wait for
anything else and will allow apps using it immediately. This is by design;
it ensures that IPv4-only networks are as fast as they possibly can be.
Therefore, this signal cannot be in any IPv6 packet, including an RA, a RFC
7050 DNS reply, a PCP NAT64 option, or a DHCPv6 option.