Re: “DHCPv6, SLAAC, Static Day X - 17 year interoperability issue” 2nd issue

[Mark Andrews <[e-mail]>]

> On 10 Nov 2020, at 12:37, Gyan Mishra <[e-mail]> wrote:
> 
> This is a separate issue from 3GPP PDP handset segmentation issue to downstream devices.
> 
> This issue exists with any deployments where the operator would like to deploy >64 prefix length host subnets at data center or access layer and this could be an enterprise or service provider use case where the router infrastructure is statically configured so no PD here.

You do know a datacenter (or any site) is not restricted to 2^16 /64 subnets.  You can get bigger allocations if you need them.  That has always been expected.

> So in this case the operator would like to deploy >64 prefixes and has let’s say all servers are configured with static and all access hosts are configured with dhcpv6 stateful RFC 8415.  No  PD here.   So we have both static and stateful hosts on the same subnet.
> 
> So now you have SLAAC hosts that get added to that same subnet  that now don’t support static or DHCPv6 stateful let’s say Chromebook or could be any device type for example and now your are in trouble.

Well if you buy a niche product and use it outside of the design space and it doesn’t work there is no one to blame but yourself.  IPv6 devices are expected to support both SLAAC and DHCPv6.

> So now due to SLAAC not supporting longer prefixes that very real fear of host operating systems that may only support slaac you and up in an terrible interoperability situation that you have to change your prefix length for all devices back to /64 so that all devices types with the 3 different IPv6 address allocation scheme can operate on the same subnet.
> 
> So due to this major 17 year issue operators have not been able to deploy longer prefix lengths to host subnets.
> 
> This is a MAJOR problem for all operators.
> 
> 
> Hope that helps clarify the interoperability issue.
> 
> 
> Gyan
> 
> On Mon, Nov 9, 2020 at 8:06 PM Joel M. Halpern <[e-mail]> wrote:
> You say "In deployment cases where you would like to deploy longer 
> prefix length subnets".
> What problem are you facing that requires longer subnets?  I understand 
> that problems that have been raised that require delegation.  That is 
> not tied to longer or shorter, and I think is better addressed by 
> shorter prefix lengths.
> 
> Is there some other problem you face that leads to needing longer prefix 
> lengths?
> 
> Yours,
> Joel
> 
> On 11/9/2020 7:35 PM, Gyan Mishra wrote:
> > 
> > This may have gotten lost in all the threads so starting a new thread. 
> >   The issue is interoperability related issue between the 3 IPv6 
> > addressing options that had been broken for 17 years.
> > 
> > Problem Statement:
> > 
> > The main point I am trying to make is that static address configuration 
> > and DHCPV6 stateful RFC 8415, you can have any prefix length prefix.   
> > SLAAC with A flag set requires a  64 bit IID and that stems from RFC 
> > 4291 modified EUI64 mac based IID generation.  However, now with random 
> > IID generation schemes with RFC 4941 privacy extension and RFC 7217 
> > stable IID you can generate any length IID.
> > 
> > The operational issue with SLAAC not supporting any prefix length in PIO 
> > and requirement for the 64 bit boundary is that in a deployment scenario 
> > where you would like to deploy longer prefix length subnets with a mix 
> > of server hosts with static address and mix of client hosts with managed 
> > address M=1 that get their 128 bit address from the DHCPv6 server pool, 
> > the fear has always been that if a device came up on the subnet that 
> > only supports SLAAC it would not work due to the SLAAC A flag set and 64 
> > bit IID requirement.  In that case the SLAAC host would not be able to 
> > communicate with any of the devices with a longer prefix including the 
> > router.
> > 
> > So that fear of interoperability of SLAAC hosts not being able to 
> > support longer prefix lengths has prevented operators from being able to 
> > deploy subnets with longer prefix lengths, as it’s hard to predict that 
> > all hosts will be able to support static or stateful and so you may end 
> > up in a situation where a device type may only support SLAAC so then you 
> > are in trouble deploying longer prefix lengths.
> > 
> > Due to the SLAAC 64 bit IID restrictions it has prevented operators from 
> > deploying “host” subnets with >64 bit prefix lengths.
> > 
> > f you go to the “root” of the problem the root is the original IPv6 
> > specification RFC 1884 dated 1995, RFC 2373 dated 1998, RFC 3513 dated 
> > 2003 and now the present RFC 4291 dated 2006.  As soon EUI64 mac based 
> > IID become not recommendedR and obsolete the standard should have 
> > immediately updated RFC 4291 as the dependency on fixed IID is no longer 
> > as now random IID generation schema starting with RFC 4941 privacy 
> > extension dated 2007 soon became standard for all OS vendors and later 
> > RFC 7217 stable IID became an alternative option to provide a “random” IID.
> > 
> > Once random IID became mainstream in all Host Operating Systems it was 
> > at that moment that the standard should have changed to update RFC 4291 
> > to permanently remove in all RFCs any mention of 64 bit boundary.
> > 
> > So this change if I do the math is now 13 years past due.  Even if you 
> > gave a few years for host operating systems to adopt the new standard 
> > which I believe back then was fairly quickly the standard should have 
> > changed eliminating the 64 bit boundary.
> > 
> > Think of all the problems “Day 17 years” this has caused and even now 
> > all of these threads.
> > 
> > This is clearly an IETF standards issue and needs to be fixed.
> > 
> > We can’t pass the blame to operators to dole out shorter prefixes or 
> > support PD.  The IETF really needs to take onus end fix the broken standard.
> > 
> > Not going to happen for PDP as their are technical issue related to the 
> > Mobile Network Gateway to support PD.   I will try to dig up the exact 
> > reason but the network element is very different then a BNG broadband 
> > gateway which supports most all L3 features.
> > 
> > As far as 3GPP operators they are following the well documented RIPE-690 
> > which only requires allocation of /64.  The main reason mobile operators 
> > are not making shorter prefix a standard is that is overkill from their 
> > perspective as you may have many mobile handsets in a household and 
> > there  is no reason for everyone at a single location to have shorter 
> > prefixes per PDP.  When you are at honme you use your wired broadband 
> > and when are away from home on the road on 3GPP on PDP is when the 
> > segmentation comes into play to subdivide the /64 prefix to downstream 
> > devices but in a household is only needed to be provided by one of the 
> > devices.
> > Bottom line is from a 3GPP provider standpoint it does not make sense to 
> > provide a shorter prefix and I don’t think that will ever change even 
> > with 5G PDP.
> > 
> > Fixed 5G broadband which is a wired broadband replacement will follow 
> > RIPE-690 guidelines and will allocation much shorter prefixes as with 
> > network slicing and other capabilities with 5G offers the requirements 
> > exist for shorter prefixes.  Not so much for PDP.
> > 
> > 
> > 
> > 
> > https://www.ripe.net/publications/docs/ripe-690#4-2-4--considerations-for-cellular-operators 
> > <https://www.ripe.net/publications/docs/ripe-690#4-2-4--considerations-for-cellular-operators>
> > 
> > 
> >         4.2.4. Considerations for Cellular Operators
> > 
> > There is a clear exception to the rule described above when assigning 
> > prefixes in a cellular network. In this case, a /64 will need to be 
> > provided for each PDP context for cellular phones, whereas for LTE 
> > modems/routers, i.e. in the case of broadband by means of cellular 
> > access, it will still be necessary to choose a /48 or /56 in accordance 
> > with the aforementioned considerations.
> > 
> > 
> > RFC 3177 for a default /48 allocation which is obsoleted by 6177 which 
> > takes a step back and is now not making any recommendations and is 
> > putting the onus on operators to figure it out and do what they feel is 
> > best which would definitely not be one size fits all approach.
> > 
> > Please read the summary section in RFC 6177 below
> > 
> > 
> >     5 <https://tools.ietf.org/html/rfc6177#section-5>. Summary
> > 
> > 
> > 
> >     The exact choice of how much address space to assign end sites is an
> >     issue for the operational community.  The recommendation inRFC 3177  <https://tools.ietf.org/html/rfc3177>
> >     [RFC3177  <https://tools.ietf.org/html/rfc3177>] to assign /48s as a default is not a requirement of the
> >     IPv6 architecture; anything of length /64 or shorter works from a
> >     standards perspective.  However, there are important operational
> >     considerations as well, some of which are important if users are to
> >     share in the key benefit of IPv6: expanding the usable address space
> >     of the Internet.  The IETF recommends that any policy on IPv6 address
> >     assignment policy to end sites take into consideration the following:
> > 
> > 
> > Thanks
> > 
> > Gyan
> > 
> > 
> > Sent from my iPhone
> > 
> > --------------------------------------------------------------------
> > IETF IPv6 working group mailing list
> > [e-mail]
> > Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> > --------------------------------------------------------------------
> > 
> -- 
> 
> 
> Gyan Mishra
> Network Solutions Architect 
> M 301 502-1347
> 13101 Columbia Pike 
> Silver Spring, MD
> 
> --------------------------------------------------------------------
> IETF IPv6 working group mailing list
> [e-mail]
> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> --------------------------------------------------------------------

-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: [phone]              INTERNET: [e-mail]