Re: [DNSOP] Making domains work even when connectivity fails (Was: the root is not special, everybody please stop obsessing over it

[Paul Vixie <[e-mail]>]

Stephane Bortzmeyer wrote on 2019-02-15 06:34:
> On Fri, Feb 15, 2019 at 09:29:29AM -0500,
>   Bob Harold <[e-mail]> wrote
>   a message of 73 lines which said:
> 
>> I think in most solutions, if the name servers for "
>> malware-c-and-c-as-a-service.com" and "com" are both unreachable,
>> the domain should continue to resolve.  But if "com" is reachable,
>> and says " malware-c-and-c-as-a-service.com" no longer exists, it
>> should go away.

this is why serve-stale is most wrong. permission, and an agreement to 
hear a trusted NOTIFY later if the authority wants to do the work of 
keeping track of who it told things to, and the work of telling them 
when something has importantly changed (like a glue address change, a 
name server change, a key change, a signature invalidated, or malware 
removed). this is a subscription (leasing) problem, not a caching one.

> Any volunteer to write a problem statement for the "VIX.SU issue"?
> Short version: "when I'm on the same network that at least one
> authoritative name server of VIX.SU, I want this domain to work, even if there
> is zero Internet connectivity to the outside world" Longer version:
> "TODO (think of: is it automatic or not, does it require prior access
> or not, phantom domains, etc)"

just as root-level is the wrong focus, so is local-level. the reason we 
don't solve this problem with multicast NOTIFY is that the information 
you may need a subscription to (due to potential network partitioning) 
could be in another campus, or another region, or another isp, or 
another country.

-- 
P Vixie