Re: [v6ops] Implementation Status of PREF64

["Pascal Thubert (pthubert)" <[e-mail]>]

I have the same perception of that thread and the IPv6 woes one on NANOG as you do. For one thing we cannot extract DHCPv6 brutally as a rotten tooth.
We need to understand what it’s used for and provide a better replacement that will slowly take over. SLAAC cannot be that. The problem of SLAAC is not the AAC piece, it’s the SL. SL lacks observability, and that’s what prevent it from being a replacement to the database that the operators and network managers need, e.g., for forensics.

I think there’s a golden path where DHCP is pushed deeper in the infra so that at first it is no more visible to the end device but still available to the operations that require it, and then we move on to the point where it becomes fully redundant and ceases to exist, by the user’s own decision, on his own agenda.

The network beyond the first hop has evolved tremendously, and we want to keep the host agnostic for the most part. RFC 8505 goes a lot way to signal things that the modern back ends (wireless, overlays) need to routing and proxying.

I just published how it can replace MLD to avoid the related multicast (https://datatracker.ietf.org/doc/draft-thubert-6lo-multicast-registration/) and that request actually came from outside the IETF and real world deployments. Note that RFC 8505 does not use MLD as it does not use the SNMA, but MLD is still a source of multicast that can be saved on the link and complexity in the stack.

If we need more, I’m happy to work on the more complete set with you all.

Keep safe;

Pascal

From: Lorenzo Colitti <lorenzo=[e-mail]>
Sent: jeudi 30 septembre 2021 11:16
To: Pascal Thubert (pthubert) <[e-mail]>
Cc: Vasilenko Eduard <[e-mail]>; v6ops list <[e-mail]>; David Farmer <[e-mail]>
Subject: Re: [v6ops] Implementation Status of PREF64

Pascal,

From what's been said so far on this thread, do you think that an implementation would achieve anything? Many of the posts here say things like, "my network, my rules", and "this network has a policy of requiring DHCPv6". Would be interested in seeing whether any of the folks on this thread who are saying that Android should implement DHCPv6 support your proposal, since it's obviously not DHCPv6. :-)

I'm all for finding another solution to this problem, but given some of the messages on this thread it doesn't look like there's much room for compromise.

Cheers,
Lorenzo

On Thu, Sep 30, 2021 at 5:43 PM Pascal Thubert (pthubert) <pthubert=[e-mail]<mailto:[e-mail]>> wrote:
There is, Lorenzo,

and strangely enough to me you are still opposing the technical evolution of SLAAC that would make them to be fully efficient – RFC 8505<https://datatracker.ietf.org/doc/html/rfc8505>.

I see that our support of First Hop Security (that includes snooping) is explicitly cited in that article. RFC 8505 solves the corner case of snooping, e.g., silent nodes which the article inelegantly ignores but are a real issue when you do not have DHCP to provide a complete state.

If needed the infra could easily republish an RFC 8505 registration to that resurrected draft-ietf-dhc-addr-registration<https://datatracker.ietf.org/doc/draft-ietf-dhc-addr-registration/> that you suggest as we do for LISP today, but we foresee a more distributed registrar, e.g., with eVPN (draft-thubert-bess-secure-evpn-mac-signaling)<https://datatracker.ietf.org/doc/html/draft-thubert-bess-secure-evpn-mac-signaling-00>.

RFC 8505 allows the device to configure any address it likes as long as it’s not duplicate. It is an alternative from DHCP where the host is still in control of its addresses; it’s still autoconf, but made stateful. It is less work on the host that already has SLAAC than implementing draft-ietf-dhc-addr-registration<https://datatracker.ietf.org/doc/draft-ietf-dhc-addr-registration/> as you suggest in you other mail.

I’m still baffled and sad that we are not working together on making this happen in a demo.

Keep safe;

Pascal


From: v6ops <[e-mail]<mailto:[e-mail]>> On Behalf Of Lorenzo Colitti
Sent: jeudi 30 septembre 2021 9:17
To: Vasilenko Eduard <[e-mail]<mailto:[e-mail]>>
Cc: v6ops list <[e-mail]<mailto:[e-mail]>>; David Farmer <farmer=[e-mail]<mailto:[e-mail]>>
Subject: Re: [v6ops] Implementation Status of PREF64

There are already vendor solutions.

https://theinternetprotocolblog.wordpress.com/2020/03/14/does-one-need-dhcpv6/

On Thu, Sep 30, 2021 at 4:12 PM Vasilenko Eduard <[e-mail]<mailto:[e-mail]>> wrote:
+1.
“Show me another solution” is a good message. Just idea or theory is not enough.
David has mentioned OpenSource. I would say that vendor product is needed too.
Ed/
From: v6ops [mailto:[e-mail]<mailto:[e-mail]>] On Behalf Of David Farmer
Sent: Thursday, September 30, 2021 5:15 AM
To: Mark Smith <[e-mail]<mailto:[e-mail]>>; Lorenzo Colitti <[e-mail]<mailto:[e-mail]>>
Cc: v6ops list <[e-mail]<mailto:[e-mail]>>
Subject: Re: [v6ops] Implementation Status of PREF64


On Wed, Sep 29, 2021 at 5:16 PM Mark Smith <[e-mail]<mailto:[e-mail]>> wrote:

On Thu, 30 Sep 2021, 03:41 Nick Hilliard, <[e-mail]<mailto:[e-mail]>> wrote:

Even if you had, that would be fine and you're welcome to your opinions.
  Other people disagree because it doesn't make sense on their deployments.

If they want to hobble IPv6, such that it is nothing more than a copy of IPv4 with bigger addresses, what is the point of going to the expense and effort of deploying IPv6 when most enterprises have plenty of IPv4 address space via RFC1918 and 100.64/10 if they were willing to abuse it a bit?

A hobbled deployment of IPv6, hobbled such that it doesn't provide any useful benefit over IPv4, is just pure business expense. Increased profit is an exceptionally strong disincentive to incurring those.

So, instead of just telling people they are doing IPv6 wrong (building a hobbled network) and that DHCP doesn't provide them what they think it does; How about making sure there are good open-source tools to build what you think is a non-hobbled network that meets their needs? In other words, how about providing some good open-source ARP and ND router scraping tools?

Now you could point the finger back at me too, but then I'm not saying that building networks with DHCPv6 is building a hobbled network, nor am I refusing to provide a DHCPv6 client for a very popular mobile and IoT platform. So, at least in my opinion, that puts more onus on you than me.

So, I agree that DHCP logging (both IPv4 and IPv6) by itself isn't enough, and yes you also need to scrape ARP and ND out of the routers. However, ARP and ND scrapping by themselves aren't enough either, DHCP logging provides much better granularity than is practical from ARP and ND scrapping, at least using SNMP. Also, by having both you can make some assumptions about suspicious access clients that are statically configuring addresses instead of doing DHCP on the access network as they should be.

I agree that limiting DHCPv6 clients to only IA-NA  and not providing IA-TA is a bad implementation of DHCPv6. Further, I recommend SLAAC, and we provide SLAAC, for general-purpose (AKA public) access networks with IPv6. But, we also have many networks where that is not appropriate, where I have regulatory and contractual compliance requirements, to protect non-public information, things like FERPA, HIPPA, PCI, and CMMC(1-4). Long-term we want these networks doing IPv6 too.

Android smartphones, probably belong on a general-purpose access network with SLAAC for IPv6 in most cases. However, Android is also on many IoT devices, things like point-of-sale terminals, credit card terminals, environmental monitoring sensors, etc... Many of those things I don't want on general-purpose access networks and some of those will have compliance requirements we have to meet. We think DHCPv6 is perfectly appropriate for these networks, and probably for server networks too.

In conclusion, while I agree with most of your arguments that DHCPv6 isn't necessarily the right way to do IPv6, especially for general-purpose (public) access networks, that doesn’t mean I think DHCPv6 doesn’t have a place in many other networks, and it would be very helpful if Android provided a DHCPv6 client, even as a non-default option.

Thanks


--
===============================================
David Farmer               Email:[e-mail]<mailto:[e-mail]>
Networking & Telecommunication Services
Office of Information Technology
University of Minnesota
2218 University Ave SE        Phone: [phone]
Minneapolis, MN 55414-3029   Cell: 612-812-9952
===============================================