Re: [dnsext] request to adopt draft-andrews-dnsext-expire-00.txt

[Kevin Darcy <[e-mail]>]

Mark Andrews wrote:
> In message <[e-mail]>, Peter Koch writes:
>   
>> On Thu, Nov 20, 2008 at 03:52:59AM +1100, Mark Andrews wrote:
>>     
>>> 	As per the chairs' instructions, this is a request for
>>> 	dnsext to adopt draft-andrews-dnsext-expire-00.txt as a wg
>>> 	item.
>>>       
>> this draft seems to assume that EXPIRE is a tool to deterministically bring
>> a zone out of service, controlled by and to the benefit of the zone
>> maintainer.  My reading of RFC 1034, section 4.3.5 is different: EXPIRE
>> is just a relief for the slave, so it doesn't have to waste more resources
>> on SOA checks.  I do not see how self-sustaining XFR-loops would hurt there.
>> Actually, many zones use dangerously low EXPIRE values and accelerating
>> expiration (independent of loops) might be counterproductive for them.
>>     
>
> 	The purpose of expire is to stop stale data being served.
>
> 	"If the secondary finds it impossible to perform a serial
> 	check for the EXPIRE interval, it must assume that its copy
> 	of the zone is obsolete an discard it."
>
> 	No where do I see any text which says that a slave will
> 	stop attempting to transfer the zone if it's old copy has
> 	expired.
>   
Indeed, unless one interprets "discard" in the RFC as implying actual 
zone removal. I interpret it as more of a "defined but disabled" status.

One would hope, however, that implementations, if any, which continue to 
attempt refreshes after expiration, would be reasonable and prioritize 
refreshes of "live" zones over ones which are expired and -might- become 
available again. Maybe an exponential backoff to some baseline, largish 
refresh interval (e.g. a day or more) would be appropriate.

                                                                         
               - Kevin


--
to unsubscribe send a message to [e-mail] with
the word 'unsubscribe' in a single line as the message text body.
archive: <http://ops.ietf.org/lists/namedroppers/>