Re: [Snac] Should a stub router learn RA header parameters from other routers?

[Esko Dijk <[e-mail]>]

Taking one step back – I just tried to summarize SNAC stub router scenario archetypes; both good and bad. Of the below it seems that only scenario 4 is now discussed as having a risk of breaking the network, correct?

1) Non-expert home user connects stub router to own unmanaged home network; it magically works and provides IPv6 stub/AIL routing – no matter whether user’s ISP offers IPv6 or not.
2) Ignorant employee brings stub router to work and plugs it into well-managed network; companies’ IT network magically protects the network from ill effects (RA guard, and/or port-based access control, … )
3) Ignorant employee brings stub router to work and plugs it into unmanaged (IPv6) network; nothing bad happens - there’s already a PIO prefix with ‘A’ on the AIL
4) Ignorant employee brings stub router to work and plugs it into not-so-well managed DHCPv6-only network; stub router sends RA PIO; many hosts on the link now autoconfigure a ULA which somehow breaks connectivity for these existing hosts.
5) Ignorant network admin connects stub router to company network without changing the default settings; then it becomes similar to 4.  This case is btw probably out of our scope to solve and could be solved by the mantra “don’t do that then”.

(The stub router is only intended for case 1, but of course the stub router cannot tell apart cases 1,4,5.)

+1 on clearly defining the possible breakage scenarios. E.g. some devices may not configure multiple addresses and get stuck with a ULA only? Or known issues with source address selection; IPv4 interactions, etc. in presence of ULAs.
When doing this the Type A, B and C IPv6 hosts may also be considered. If we find no issues then we may also state this in the draft in an Appendix – what was considered, and why it isn’t a problem.

Esko

From: Lorenzo Colitti <[e-mail]>
Sent: Friday, September 8, 2023 11:38
To: Ole Trøan <[e-mail]>
Cc: Lorenzo Colitti <lorenzo=[e-mail]>; Ted Lemon <[e-mail]>; Juliusz Chroboczek <[e-mail]>; Esko Dijk <[e-mail]>; Michael Richardson <[e-mail]>; Handa Wang <[e-mail]>; [e-mail]
Subject: Re: [Snac] Should a stub router learn RA header parameters from other routers?

On Fri, Sep 8, 2023 at 5:28 PM Ole Trøan <[e-mail]<mailto:[e-mail]>> wrote:
I am not OK with a SNAC router connecting to an existing IPv6 network and breaking it.

If you can elaborate on the type of breakage you are concerned about, then we can discuss it and possibly design for it.

FWIW, just the fact of the hosts getting new addresses does not qualify as breakage IMO. All IPv6 hosts have multiple addresses and source address selection generally gets hosts to do the right thing.