Re: [v6ops] SLAAC renum: Problem Statement & Operational workarounds

[Ted Lemon <[e-mail]>]

On Oct 30, 2019, at 10:03 AM, Bernie Volz (volz) <[e-mail]> wrote:
> A Solicit kind of means “start from scratch”. So that’s why the current RFC8415 specification does not require sending information related to “old” leases. The text also does not provide for sending 0 lifetimes in an Advertise which is where this would first appear – and then also in the Reply to the Request. Perhaps there is an argument that these leases should not be added to the Advertise, but be in the Reply to the Request.
>  
> So, it does get a bit more complex and needs to be worked out.
>  
> > I don’t see any reason why this should not just be the default behavior.
>  
> As I said that as this is a change in behavior, we don’t know what the consequences for existing devices may be – and in some cases, if the device does have storage, would this confuse it? I’m not against having this be a configuration knob in the server which can be applied at various “levels” (perhaps even globally).

There are a lot of assumptions you’ve stated here that as far as I know do not appear in the standard.   E.g., you may think that a solicit kind of means “start from scratch,” but is it not okay for the DHCP server to send 0 lifetimes in the Advertise?  Or is it just not suggested?

I agree that it needs to be worked out.   As for existing devices, I think that at this point a scorched earth policy is not unreasonable: if there are devices out there that have firmware that can’t be updated to fix problems like this, it’s probably good if they stop working and have to be “recycled.”   I hope that number is zero; if for some ISP it’s sufficiently more than zero that this isn’t the right solution, then they may have to wait until those routers EOL before deploying this “new" behavior.

However, I think in practice this simply wouldn’t cause a problem, and complexifying things to prevent such a problem seems unnecessary in the absence of data.

Do we test for the case where the DHCP server sends back two IA Prefix options in an IA_PD?   I think that’s allowed, isn’t it?   Maybe the right behavior is to leave the Advertise unchanged, but send to IA Prefixes in the Reply.