[dns-operations] Reverse Lookup speedup

[Stephane Bortzmeyer]

On Thu, Aug 02, 2007 at 08:09:38AM +0200,
 Michael Monnerie <michael.monnerie at it-management.at> wrote 
 a message of 57 lines which said:

> I've very interestingly read the thread about NetBSDs new root
> slaving,

Do note it is FreeBSD, not NetBSD.

> and that some are pissed off by this, and grabbed the hint that
> loading the root zone from ftp.internic.net is also possible and I
> immediately did this for one heavy loaded DNS server.

Bad idea: "slaving" the root from ftp.rs.internic.net is less work for
the root name servers, true, but it is also more brittle because,
instead of using the proved and tested mechanism of SOA queries + AXFR
transfers, you rely on a homemade script and the good will of
cron. The risk of becoming stale if one of these components fail is
too high, IMHO.

> I was wondering if there's some similar way of loading (or slaving)
> a reverse lookup root zone.

FreeBSD new named.conf does it, too.

> I still do not understand how reverse lookups are done. Is that just
> the .in-addr.arpa zone that would need to be loaded, and if yes,
> where could I automatically get it from?

[I suggest that you may delay introducing invasive changes like this
one, if you do not know the DNS in detail.]

Slaving in-addr.arpa, like FreeBSD does, does not save you a lot of
lookups. Unlike forward requests, where there are a lot of typos which
can be catched only by the root, reverse lookups always have an
existing "TLD" and therefore I suspect that the cache is sufficient.