Re: Objection to draft-ietf-6man-rfc4291bis-07.txt

[Alexandre Petrescu <[e-mail]>]

Le 01/03/2017 à 12:04, [e-mail] a écrit :
>
>> On 1 Mar 2017, at 11:58, [e-mail] wrote:
>>
>>>>> FWIW I wouldn't oppose an exception for manually-configured
>>>>> addresses on routers.
>>>>
>>>> But you would oppose it for hosts? Then I don't think we're
>>>> any further.
>>>>
>>>
>>> Not sure about that. I suppose it also depends on what manual
>>> means. If it's literally just operators configuring them manually
>>> by editing configuration files that's probably fine, but that's
>>> not what you want to do, is it?
>>
>> Yes, that is exactly what I want to do. I want to be able to use
>> the ifconfig command to set an IPv6 address with a netmask
>> different from 64, and to have the corresponding config file entry
>> take effect when the box is booted or the service restarted.
>
> And that clearly always have been permitted.

In a sense yes, but in the sense of the longest-match algorithm this
permission is not that clearcut, because it is over-ridden.

Let me explain an example.

I am permitted to freely add a fe80::1/63 on my linux but the
longest-match algorithm will over-ride my fe80::/63 intention because
the kernel over-rides it with a _by_default_ fe80::/64 prefix it added
there without me asking it.  That _by_default_ is also resisting reboots.

So - yes, it is permitted, but only to be over-ridden by the 64.  In
other words: I am permitted to add a fe80::1/63 and I will be succsefull
in my intention _if_ I also delete the kernel's intention which is there
by default, and that after each reboot.

(in Windows there is a similar thing).

In that sense, I agree with you.

> E.g. how would you otherwise configure a /128 loopback.

In linux /128 is a particular case.  It pops up after an ifconfig if
that ifconfig did not specify a plen parameter.

There are /128s on other interfaces than loopbacks.

There may be a confusion in considering that /128 is a prefix length, or
it just means an address.  DHCPv6 people think a /128 means just one
address, whereas RA people think /128 is something that could be put in
an RA.

> I agree that current text have always been at best ambiguous on this
> point.

I agree.

Alex

>
> Best regards, Ole
>
>
>
> --------------------------------------------------------------------
> IETF IPv6 working group mailing list [e-mail] Administrative
> Requests: https://www.ietf.org/mailman/listinfo/ipv6
> --------------------------------------------------------------------
>