DNS tutor needed

[Simon Waters]

Barry Margolin wrote:
> 
> >> That's negative TTL not default.
> >
> >Sorry?
> 
> Since BIND 8, the last field in the SOA record has been the TTL of negative
> caching, not the default TTL.  The $TTL directive sets the default TTL.

And as I have pointed out before, BIND still defaults to using
the last field of the SOA as a default TTL if no $TTL directive
exists.

This appears to contradict the RFC that says NOT TO USE the last
field of the SOA as the default TTL. 

It is a minor point, but the RFC appears to prefer a default
like 86400 seconds for default TTL, over the actual behaviour of
BIND. The main result is many people still overload the last
field of the SOA, and use either an inappropriately high
negative TTL, or an inappropriately low default TTL.

If BIND was modified to supply a more sensible default
behaviour, we could lift the restriction on the negative TTL
maximum from 3 hours, drop in a warning message if the last
field of the SOA is more than say 3 hours, and probably save
upgrading a few root servers, and make the whole Internet a tad
more responsive in the process.

Since it won't end world poverty, or anything, it probably won't
happen, but given how aggressive applications are becoming in
placing more DNS queries (IPv6, Dynamic DNS etc, DNSSEC,
opportunistic encryption), then appropriate negative caching is
going to become more important.