Re: you have running code ... I-D Action: draft-ietf-6man-ipv6only-flag-03.txt

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

On 30 Oct 2018, at 20:27, Job Snijders wrote:

Hi,

> It is great to see people produce running code for experimentation -
> congratulations!
>
> Can you (or others running FreeBSD EXPERIMENTAL) share reports on how
> this pans out in practise?

Well, it would be great if I could turn the flag on on various other 
default routers on the networks I am normally on.  Hope $C and $J and $A 
at least will have this one shipping quickly once it’s an RFC.


> What gaps between reality and documentation
> are uncovered,

I’ve shared little notes with Bob and Brian and I am sure they’ll 
update things;  I’ll send some of that out there the next days (if 
they don’t beat me).  I also patched tcpdump to see the flag and I can 
share some ifconfig/ndp output if it’s desirable.


> or whether the patch was complete/not-complete?

Software is never complete ;-)   As I said “implement a minimalistic 
set of this”.  That’s what it is.
If I would add to this (and I am sure I will) it’ll be the following 
(in order):
(a) RX filter based on interface flag so packets don’t make it up the 
stack at least anymore
(b) a global (system-wide, not per-link) flag to allow an administrator 
to overrule the decision (for whatever reason)
(c) allowing the system admin to set the interface specific IPV6_ONLY 
flag in a sticky way manually without the RAs (hey I am fine with 
v6only, don’t try v4 even if you could);   also great for apps people 
to test if they can’t get a full test network to play with.
(d) deal with dhclient and do make ipv6 start before ipv4 as indicated 
in the draft.


I could go and teach gazillions of user space applications about it 
(dhclient probably one of the most nagging ones) to save local battery 
cycles as well, but then I could just compile a FreeBSD client without 
INET support (which we did in 2011) and be done with it.   I think what 
I am trying to say:  why waste time on software which is almost 
obsolete?

I’ve done an experiment years ago, running an “IPv6-only” kernel 
with EAFNOSUPPORT coming back earlier in the stack rather than later.  
Also we have a library function in FreeBSD which indicates if the 
feature “inet” or “inet6” (and others) are present or compiled 
out.  So it’s been there for apps people to test with for 7+ years..  
=this is just yet another flag to care about.


> What OS vendor will go next?

All hopefully :-)

>
> Kind regards,
>
> Job
>
> On Tue, Oct 30, 2018 at 9:13 PM Bjoern A. Zeeb
> <[e-mail]> wrote:
>>
>> On 16 Oct 2018, at 23:09, [e-mail] wrote:
>>
>>> A New Internet-Draft is available from the on-line Internet-Drafts
>>> directories.
>>> This draft is a work item of the IPv6 Maintenance WG of the IETF.
>>>
>>>         Title           : IPv6 Router Advertisement IPv6-Only Flag
>>>         Authors         : Robert M. Hinden
>>>                           Brian Carpenter
>>>       Filename        : draft-ietf-6man-ipv6only-flag-03.txt
>>>       Pages           : 12
>>>       Date            : 2018-10-16
>>>
>>> Abstract:
>>>    This document specifies a Router Advertisement Flag to indicate 
>>> to
>>>    hosts that the administrator has configured the router to 
>>> advertise
>>>    that the link is IPv6-Only.  This document updates RFC5175.
>>
>> You have running code:
>>
>> https://lists.freebsd.org/pipermail/svn-src-head/2018-October/119360.html
>>
>> I wish the discussions would be as short as the code changes that are
>> needed to implement a minimalistic set of this.
>>
>> Love and peace and no more IPv4 noise.
>>
>> /bz