Which TTL will bind use ?

Kevin Darcy kcd at daimlerchrysler.com
Thu Apr 18 21:44:04 UTC 2002


Torsten Mueller wrote:

> Kevin Darcy schrieb:
> >
> > Torsten Mueller wrote:
> >
> > > I have a question regarding the TTLs.
> > >
> > > Assume that you have an A record for
> > > test.mydomain.org with the TTL 86400 and the IP
> > > 192.168.1.1
> > >
> > > Now you try to add via dynamic dns update a second A record
> > > with a different TTL i.e.
> > > test.mydomain.org with TTL 20000 and the IP
> > > 192.168.33.10
> > >
> > > I know it's "illegal" and bind only accepts only 1 TTL for
> > > the same host with the same record type.
> > >
> > > My question is, which of the TTLs bind will use for the 2 records ?
> > >
> > > The TTL with the highest value ?
> > > The TTL which was "the first" TTL ?
> > > The TTL of the last submitted record ?
> > > one of the TTLs (random decision of bind) ?
> >
> > According to RFC 2181, Section 5.2, nameservers cannot *send* an RRset
> > with differing TTLs. If they *receive* such an RRset, then they
> > "minimize" the TTLs to the lowest TTL value. This also appears to be
> > what BIND does when receiving a Dynamic Update which causes differing
> > TTLs in an RRset (I just tested it).
>
> Thanks Kevin,
> i took for generation of this "irregular" TTL the dnszone perl module (a
> webinterface
> for managing zones via [signed] Dynamic Updates) , which uses Net::DNS.
>
> My experience however is , that bind uses the last submitted TTL.
> I tested this with bind 9.2.0
>
> Which version did you test against ?

Actually, I tested with a BIND 8 nsupdate and nameserver. Maybe it behaves
differently. As far as I can tell, the correct behavior isn't clearly
specified by any RFC.


- Kevin





More information about the bind-users mailing list