Re: Confirmation to advance: draft-ietf-6man-ipv6only-flag-05

["Bjoern A. Zeeb" <[e-mail]>]

On 24 May 2019, at 14:47, Philip Homburg wrote:

>> I wonder if the ipv6only flag could be used to test IPv6 on networks.
>> That is, turn off IPv4 and see if the IPv6 is working.  Turn off 
>> IPv4,
>> then make sure there is working IPv6 connectivity, throughput, etc.,
>> then turn IPv4 back on.
>>
>> Perhaps a new use case.
>
> To make that work you have to turn off DHCPv4 as well. So you don't 
> need
> the flag.

That wasn’t the case with the earlier initial versions of the draft; 
the ability for that “use-case” only went away with people asking 
for it as they are worried about legacy accidentally turned off by a 
rouge RA packet.

There’s still good ways to use the two options together for testing if 
I think about tuning lifetimes of each of these protocols, etc.

It was all simpler when the draft was simple and I wonder if with the 
(global) flag to turn off enabling of processing the flag, the “IPv4 
address configured on the interface” requirement could be removed 
again from the draft;  that would also avoid cross-protocol-family 
pollutions in the implementations.


In addition to that, I challenge you to see what a network without 
DHCPv4 looks like when you check the “legacy IP noise levels” (even 
if there is no static IPv4 configurations and even more so if you’d 
only send DHCPv4:116).


> The bigger problem is that without IPv4, most of the internet is 
> unreachable.
> So it is bit hard to test if it is IPv6 that is failing or not.

You can still test (a) that your local services on the LAN work with 
IPv6-only, that you can reach (b) important (external) IPv6 enabled 
services, (c) if anything breaks if you go through a proxy or 
translation service, .. (just to name the first three which came to my 
mind).


/bz