To:
"D. J. Bernstein" <djb@cr.yp.to>
cc:
dnsop@cafax.se
From:
Shane Kerr <shane@ripe.net>
Date:
Wed, 30 May 2001 11:03:36 +0200 (CEST)
In-Reply-To:
<20010529232301.5659.qmail@cr.yp.to>
Sender:
owner-dnsop@cafax.se
Subject:
Re: Should a nameserver know about itself?
On 29 May 2001, D. J. Bernstein wrote: > Unfortunately, neither ARIN nor RIPE supports glue, despite the last > paragraph of RFC 1034 section 4.2.2, so reverse lookups on the > Internet are unnecessarily rickety. Remember that registries are concerned with address to name mapping, and this is a name to address mapping - one that is already performed elsewhere. The problem as I see it is that in order to provide glue records, the RIR's need to track the A records similar to what you demonstrated. What this means is that when an in-addr.arpa zone changes (e.g. new nameserver added, nameserver IP changes) the zone administrator has to remember to update both their own zone files (or equivalent in the tinydns case) as well as the records at the appropriate RIR. One possible solution is for the RIR's to allow the recipients of delegated space the option of maintaining glue information, so they can decide for themselves whether they want the administrative overhead in exchange for increased reliablity and response, reduced network and server load, etc. In reality it is likely there would be a lot of garbage - perhaps even reducing the reliability of IN-ADDR.ARPA in general rather than increasing it. Would the RIR's then have to run automatic processes to verify their glue records? Would (should) they then simply use those all the time rather than requiring users update them manually at all? There are large disadvantages to an automatic process like this. IIRC, I asked for community input about this issue when I was at ARIN, and received no feedback - my guess is that the Internet at large doesn't consider it an important issue. -- Shane