RE: SLAAC vs DHCPv6 (II)

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

Exactly!

Also, I hope this discussion does not turn into a router vendor against a host vendor for who takes the blame in case of an outage. The real question is what gets us the best IPv6 service for the end user and the network admin. 

As you point out, this discussion has already isolated environments, e.g., home when all must work automagically but the multicast pressure is low, and corporate/marge conference where there's a control center and a need for visibility. 

For home, there are a limited number of devices and if the hosts respect whatever best practice, there should not be a problem in normal conditions. There can be misbehavior on the host side, leaks on the router side. I guess the router would malloc to death and reboot when saturated, and at home that would be OK if rare enough. So I just expect "no limits" by default in the home GW. Which mean that the router would automagically have a state maintained for the user addresses for their requested lifetimes, with a better user experience when the address is allocated - Jen's point-, but not much difference.

In a managed environment things are different. If a NOC configures the router to say "sorry, network is full" when the host has one address, and the users complain, it will be between the users and the NOC, not between the router and the host. The network, devices and protocols are not to blame. The NOC can already set that limit today on a variety of products. But for the lack of explicit signaling, the addresses that the network maintains may not be the ones the device wants to use.

On the SAVI point, RFC 8505 makes SAVI more efficient by maintaining a real state as opposed to snooping protocols. Current SAVI experience shows limitations with silent hosts (e.g., a printer that does not expose an address, did DAD last year and still expects to be reachable on that address), mobility (hard to differentiate with impersonation attack), and others. See also with https://datatracker.ietf.org/doc/draft-ietf-6lo-ap-nd/ how it is a lot easier to do a SEND equivalent with a registration.

Cheers;

Pascal


> -----Original Message-----
> From: ipv6 <[e-mail]> On Behalf Of Nick Hilliard
> Sent: mardi 28 janvier 2020 14:39
> To: Lorenzo Colitti <lorenzo=[e-mail]>
> Cc: 6man WG <[e-mail]>
> Subject: Re: SLAAC vs DHCPv6 (II)
> 
> Lorenzo Colitti wrote on 28/01/2020 10:14:
> > Second: for many networks, the desired saturation point is one
> > address. I think we here all agree that is a bad idea
> 
> This seems to be a curiously simplistic view of a complex policy issue for which
> there is no single answer.
> 
> Specifically we don't all agree that this is a bad idea and many people in this
> working group and others have been very clear that whether this is a good or a
> bad idea depends on the situation.
> 
> Nick
> 
> --------------------------------------------------------------------
> IETF IPv6 working group mailing list
> [e-mail]
> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> --------------------------------------------------------------------