Difference in SOA record between master and slave?

[Jesper Dybdal]

Barry Margolin <barmar at bbnplanet.com> wrote:

>In article <38B67F4D.818.46F3E26 at localhost>,
>Richard Stevenson <RichardS at webmasters.co.nz> wrote:
>>My initial thought was that the zone file wasn't being transferred  
>>correctly, but that's not the case (I've pasted in both files below - from  
>>the master and from the slave).  It looks as if ns1.akl.adv.net.nz is  
>>ignoring the TTL specified on the SOA record in favour of the minimum TTL  
>>in the SOA record itself.  Is this normal? 
>
>The specification of negative caching says that the SOA records's TTL
>should be set to the minimum of the MinTTL and the TTL specified explicitly
>for the SOA record.  I suspect this is because early drafts of the ncache
>specification said to use the SOA record's TTL as the ncache TTL.  So this
>ensures that servers that implement that earlier scheme will use the proper
>ncache TTL when they're talking to newer servers.

But the strange thing is that the master server does not do this: it provides
the full specified (or default) TTL for the SOA record, in a response as well
as in a zone transfer.

The slave server receives the zone transfer with the full TTL and saves it in
the file with the full TTL, but when it replies to requests it inserts the
small minimum/ncache TTL as the TTL for the SOA record instead.

I'm using 8.2.2-P5 for the master as well as for the slave.
-- 
Jesper Dybdal, Denmark.
http://www.dybdal.dk (in Danish).