Re: [v6ops] cpe-slaac-renum: Proposed text for prefix lifetimes

[[e-mail]]

Ted,

>> No, that's why the delegator needs to remove those prefixes properly. I.e. by including them in DHCP with lifetime of 0.
>> In the flash renumbering case, the delegator will set the lifetime to 0 from whatever it was before. No need to "wait out" the lifetime.
>> 
>> Long valid lifetimes affect what you put in DNS as TTL for the record for example.
> 
> Okay, but you are talking about a very different operational use case.  This is how you would do things in a data center.
> 
> In the case of a CPE router, it’s not safe to assume the router won’t be unplugged or power cycled.  You probably don’t _want_ to use long TTLs on domain names, if any are published—e.g., for mDNS we use a TTL of 75 minutes, and do a refresh at 60 minutes; we do the same thing for Service Registration Protocol (DNS updates for DNSSD), although that’s not widely deployed at present.
> 
> So in this case, where your device may be power cycled and lose its state, or where the end user may simply unplug one device and plug in another, having a really long preferred lifetime makes zero sense.  Obviously, having a shorter lifetime does not solve the problem of cleaning up stale prefixes fast enough to prevent routing problems, but we still want those stale prefixes to be cleaned up.  As you say, there are two different issues here.  Reasonably short address lifetimes avoids keeping addresses around for a month when they are stale and aren’t going to be explicitly deprecated.  For source address selection and route selection, this doesn’t help, but it’s still worth doing.
> 

You seem to be implying something here.
Why does it matter if a CPE is unplugged or power cycled? (Well, apart from the obvious of being unable to forward packets for the duration).

Ole