Re: IPv4 traffic on "ietf-v6ONLY"

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

On 16/11/2017 22:44, Tim Chown wrote:
>> On 16 Nov 2017, at 00:56, Brian E Carpenter <[e-mail]> wrote:
>>
>> On 16/11/2017 13:39, David Farmer wrote:
>>> On Wed, Nov 15, 2017 at 5:54 PM, Brian E Carpenter <
>>> [e-mail]> wrote:
>>>
>>>> On 16/11/2017 08:15, james woodyatt wrote:
>>>>> On Nov 15, 2017, at 02:47, Philip Homburg <[e-mail].
>>>> com> wrote:
>>>>>>
>>>>>> The safest option to do that is a DHCPv4 option that says 'no IPv4
>>>> service here, go away'.
>>>>>
>>>>>
>>>>> Better: extend ARP with a signal that says, “ARP is not welcome here."
>>>>
>>>> However, the IPv4 traffic seen on ietf-dns64 is negligible and harmless.
>>>> That is a practical indication that this problem probably isn't worth
>>>> solving.
>>>>
>>>> I submit for example that sending a new "not welcome" response to
>>>> ARP requests would do more harm than good, since the legacy hosts
>>>> would probably react badly.
>>>>
>>>>   Brian
>>>>
>>>
>>> I'd love to see some data to back up the conjecture, that is the "traffic
>>> ... is negligible and harmless"
>>
>> That it's negligible was shown in Bob Hinden's statistics from Wireshark.
>> That it's harmless seems obvious - it's virtually all multicast discovery
>> traffic that goes nowhere. As I said, I have a couple of capture files
>> if anybody is interested.
> 
> Careful with that ethics can of worms you’re holding there ;)

Yes, maybe we need to add "IP packets" to the last sentence of the Note Well ;-)

   Brian
 
> I’ve seen some university networks where the wireless is surprisingly flat, and the multicast/broadcast traffic is not insignificant. The more recent trend is to VLAN pool or similar to avoid that, but that then affects service discovery (two WiFi devices in same room, but potentially on different subnets).> 
> The service discovery traffic is generally replicated over IPv4 and IPv6; 
> you’ll see the same advertisements and discovery traffic essentially duplicated.  But this is potentially useful in say a home network where you have legacy devices like TVs for which you can only discover the display over IPv4, and throw data to it over IPv4. 
> 
> The model in the DNSSD WG for wider area discovery is to keep resolution local, and relay the responses from the discovery proxies using unicast. 
> 
> Tim
>