bind-9.8.1: INSIST(! dns_rdataset _isassociated(sigrdataset)) failed

[David Ford]

ISC have replied and indicated that BIND 10 was designed, with
resilience to abnormal events, in mind.  i'm eagerly looking forward to
trying it out now.

i disagree that it's easier to find and fix.  many people will simply
wrap it in a while(1) and ignore it because we don't have the time to
sit and debug it, which puts us precisely at the same footing as simply
logging an alert and continuing -- except we now have the lag during the
restart period.  events often occur at inopportune moments.

-d

On 11/16/2011 11:47 AM, Paul Wouters wrote:
> These however do guarantee internal state so any kind of new bug is
> much easier
> to find and fix. Openswan does the same thing for this very reason.
> However,
> openswan does have an init script that runs a while(1) loop over its
> daemon.
> This means once we encounter unexpected state, we drop all state and
> restart.
>
> Perhaps bind and/or distributions should also use such an init script.
> I would prefer
> that over attempting to continue with a bad internal state and seeing
> apparent
> random state/crashers later on in bind because it tried to continue
> after something
> bad.
>
> Paul
>