Re: [EXTERNAL] Re: SLAAC, Static & DHCPv6 day 1 interoperability issue

Le 09/11/2020 à 23:13, Templin (US), Fred L a écrit :
> Brian, brief comment/question below:
> 
>> -----Original Message----- From: ipv6
>> [mailto:ipv6-bounces@ietf.org] On Behalf Of Brian E Carpenter Sent:
>> Monday, November 09, 2020 1:58 PM To: Gyan Mishra
>> <hayabusagsm@gmail.com> Cc: IPv6 IPv6 List <ipv6@ietf.org>;
>> draft-mishra-6man-variable-slaac@ietf.org; Alexandre Petrescu
>> <alexandre.petrescu@gmail.com>; Dusan Mudric
>> <dusan.mudric@gmail.com>; Dmytro Shytyi <dmytro@shytyi.net> 
>> Subject: [EXTERNAL] Re: SLAAC, Static & DHCPv6 day 1
>> interoperability issue
>> 
>> This message was sent from outside of Boeing. Please do not click
>> links or open attachments unless you recognize the sender and know
>> that the content is safe.
>> 
>> 
>> In line...
>> 
>> On 10-Nov-20 04:35, Gyan Mishra wrote:
>>> Brian
>>> 
>>> In-line
>>> 
>>> On Sun, Nov 8, 2020 at 3:14 PM Brian E Carpenter
>>> <brian.e.carpenter@gmail.com
>>> <mailto:brian.e.carpenter@gmail.com>> wrote:
>>> 
>>> Gyan,
>>> 
>>> I don't think you were around for the original discussions, so
>>> there is an aspect that is missing from your logic below.
>>> 
>>> The inclusion of a separate interface identifier field in IP
>>> addresses was an entirely intentional feature of IPng. If all we
>>> had wanted
>> to do was IPv4 with bigger addresses, that's what we would have
>> done and the address length would have undoubtedly been 64 bits. In
>> fact there were various proposals to do exactly that, with a
>> variety of associated transition and coexistence mechanisms.
>>> 
>>> But the rough consensus was to do more than that, and to allow
>>> *extra* space in the address for an interface identifier that
>>> was
>> not part of the subnetting mechanism. Originally it was going to be
>> 48 bits, so the longest subnet prefix would have been 80; on second
>> thoughts it was set to 64, which gave *exactly* the same extension
>> to the subnettable space as we would have got from IPv4 with bigger
>> addresses.
>>> 
>>> That isn't inconsistent with what we now call BCP198, which says
>>> that on links where an interface identifier & SLAAC isn't
>>> needed,
>> subnetting can extend out to /127.
>>> 
>>> All that was despite the fact that we hadn't even realised the
>>> potential privacy benefits of a host-defined interface identifier
>>> at the
>> time; that is much more recent.
>>> 
>>> As far "day 1" goes, please remember that DHCPv6 is a retro-fit:
>>> 
>>> RFC1971 IPv6 Stateless Address Autoconfiguration. August 1996 
>>> RFC3315 Dynamic Host Configuration Protocol for IPv6 (DHCPv6).
>>> July 2003.
>>> 
>>> 
>>> Gyan> Makes sense then that as DHCPv6 was a retrofit “add on” to
>>> the base architecture that this issue came about afterwards.
>>> 
>>> 
>>> 
>>> (In fairness, draft-ietf-addrconf-ipv6-auto-00 was dated January
>>> 1995 and draft-ietf-dhc-dhcpv6-00 was dated February 1995, but
>> it advanced very slowly compared to SLAAC.)
>>> 
>>> 
>>> Gyan> From a problem statement perspective do you agree with the
>>> title of this thread “Day 1 interoperability issue”?
>> 
>> No. From the dates of the RFCs, it's a "Year 7 interoperability
>> issue".
>> 
>>> Do you agree that one way to solve is to allow SLAAC to support
>>> longer prefix lengths?
>> 
>> That's one way, but it's the wrong way. The right way is for all
>> operators, including mobile operators, to assign /48 or /56 to all
>> end users.
> 
> Isn't that exactly what RFC6177 (BCP157) tells us? Should we be
> working to reaffirm that that BCP still applies today?

FRed, let me say now what I said then too: that RFC is for sites, not
for mobiles.  Some explain it as being the same, but they are not the same.

An RFC for mobiles, like RFC6177, might be appropriate.  But I think one
cant make it at IETF because concerned (~) operator(s) would not let it
through.  One simply cant accept an RFC which practically outlaws that
many deployments.

The (~) operators who would accept /56 to mobiles would suddenly reject
so many other protocols and software they already installed, planned, etc.

Alex

> 
> Thanks - Fred
> 
>>> Do you agree that this is a major operational issue that needs to
>>> be solved?
>> 
>> Yes, but as Barbara says, that needs some collaboration with the
>> SDOs and operator fora to get rid of /64 assignments.
>> 
>> Brian
>> 
>> --------------------------------------------------------------------
>>
>> 
IETF IPv6 working group mailing list
>> ipv6@ietf.org Administrative Requests:
>> https://www.ietf.org/mailman/listinfo/ipv6 
>> --------------------------------------------------------------------