Re: Running code (Was: I-D Action: draft-ietf-6man-ipv6only-flag-03.txt)

[Mark Smith <[e-mail]>]

On Sun, 28 Oct 2018 at 02:02, Mikael Abrahamsson <[e-mail]> wrote:
>
> On Sat, 27 Oct 2018, Manfredi (US), Albert E wrote:
>
> > Either way, the only hosts which will NOT holler, when you shit off
> > IPv4, are the updated ones. Updated to read this flag, or updated with
> > the better heuristics, makes no big difference.
>
> For me, this flag has always been an additional input to the heuristics
> engine.
>

I don't think heuristics are really necessary.

A host doesn't really need an IP address (IPv4 or IPv6) unless and
until it starts running an application that wants to use the network.

I think the current convention for hosts to acquire addresses upon
boot or more recently upon attachment to a (new) link is probably more
of a tradition from back when minicomputers were the norm and network
interface configuration was always static. Hosts now widely support
dynamic address acquisition or generation, and release.

Hosts acquiring or generating addresses at boot or upon link
attachment could be seen as an optimisation, with the optimisation
being to avoid address acquisition latency at the time of the first
application network API call.

With IPv4 being the legacy protocol in this scenario, IPv4 addresses
could be acquired and released on demand by hosts when IPv4
applications need them.

I think on-demand IPv4 would be a compliment to an IPv6 Only flag
signalled by an IPv6 only router. There are still optional things that
Dual Stack hosts do that aren't really networked application driven,
such as responding to ICMP Echo Requests or registering host presence
via a HINFO record using Multicast DNS.

An IPv6 Only flag would be an explicit signal to a Dual Stack host
that these optional IPv4 things should be switched off. If they're the
only thing remaining that uses IPv4, then the host would cease to use
DHCPv4 to acquire an IPv4 address.

When a host stops running any applications the use IPv4, and these
other optional IPv4 hosts do are switched off, then the host would
cease sending or receiving an IPv4 or IPv4 related ARP packets
entirely.

Regards,
Mark


> We discussed this in sunset4 about exponential backoffs etc. This flag can
> just be additional inputs.
>
> But yes, I also see the possibility of the host trying to configure itself
> IPv4 but if it sees no answers to questions and doesn't see any IPv4
> traffic at all, after a while it "gives up" and just goes "IPv4 dormant"
> and sits there and monitors to see if it sees any IPv4 traffic. It might
> restart IPv4 operations when it sees other IPv4 traffic (0x0800 or
> 0x0806).
>












> --
> Mikael Abrahamsson    email: [e-mail]
>
> --------------------------------------------------------------------
> IETF IPv6 working group mailing list
> [e-mail]
> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> --------------------------------------------------------------------