IXFR and lost (?) updates of a TTL

[Chris Thompson]

On Apr 24 2006, Stephane Bortzmeyer wrote:

>On Mon, Apr 24, 2006 at 12:16:57PM +0200,
> Stephane Bortzmeyer <bortzmeyer at nic.fr> wrote 
> a message of 48 lines which said:
>
>> We experienced a curious BIND bug on slave machines which use IXFR to
>> transfer from a BIND master. The TTL on the NS records were
>> changed. While the master had:
>
>Well, thanks to Jean-Philippe Pick and Stefan Puiu, we can now
>reproduce the bug :-)
>
>If, on the master, you just change the TTL of a RRset, this change is
>not reflected in the journal file (as displayed by journalprint) and
>obviously not propagated to IXFR clients (AXFR clients - and regular
>DNS clients like dig - see the new TTL).
>
>Our master is 9.3.1. 

Could this be:

1855.   [bug]           ixfr-from-differences was failing to detect changes
                        of ttl due to dns_diff_subtract() was ignoring the ttl
                        of records.  [RT #14616]

fixed in 9.3.2? Your original post doesn't make it obvious whether you are
using ixfr-from-differences or not.

In any case, experiments with 9.3.2 seem to show that it constructs correct
IXFRs from changes of TTL (only), whether that is done via dynamic update
or via ixfr-from-differences.

-- 
Chris Thompson
Email: cet1 at cam.ac.uk