Re: [DNSOP] Fwd: New Version Notification for draft-pusateri-dnsop-update-timeout-01.txt

[Joe Abley <[e-mail]>]

On 20 Feb 2019, at 00:35, Paul Wouters <[e-mail]> wrote:

> On Wed, 20 Feb 2019, Mark Andrews wrote:
> 
>> Think disaster recovery and promoting a slave to master.  You have to
>> transfer state between servers.  You can transfer it in band or out of
>> band.  If you transfer it out of band you need to invent / specify
>> yet-another-protocol to do it on top of specifying when records need to
>> be removed.
> 
> That is not very convincing to me. disaster recovery scenarios seem to
> be solvable by proper admin and by the daemon properly writing state to
> disk which can be saved off-site. It also seems a rather rare occurance.

I agree.

The crux of the use case seems to be that it is commonplace for names in the DNS to exist for short periods of time and that for some applications a name that overstays its welcome can cause an operational problem.

While I can understand the philosophical desire to complete the UPDATE specification so that it is possible to engineer around this scenario, I don't see the practical application. The existence of the requirement in the first place seems unproven (at least there are no obvious examples given); the scenario in which the purported operational problem arises seems likely to be rare; workarounds surely exist and, really, the damage in the event that the stars align and a temporary name does persist seems very low.

If the goal is to try this out and have no impact on existing implementations (e.g. there is some side code that is imagined that will poll a transferred zone for TIMEOUT records and do local UPDATEs in order to remove RRSets that should not be there) then all that is really needed here is a code point for the TIMEOUT RR. The existence of the draft is nice since documentation is good, but I think "experimental" would be a better target than "standards track". It's surely possible that this mechanism will solve some as-yet unnoticed, large-scale problem and will one day be considered essential functionality, but I don't think we're there today. There be camels.


Joe