[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]


To: <ietf-provreg@cafax.se>, Kent Crispin <kent@songbird.com>
From: Sheer El-Showk <sheer@saraf.com>
Date: Fri, 6 Apr 2001 18:23:10 -0400 (EDT)
In-Reply-To: <20010406121036.A29000@songbird.com>
Sender: owner-ietf-provreg@cafax.se
Subject: Re: Nameserver MUST HAVE IP

> The requirement does not prevent names from becoming unresolvable --
> there are lots of other ways that can happen, and arguably, the
> increased churn in the address data leads to more misconfigurations than
> it prevents.  It's hard to say.

Personal experience has shown this to be the case.  I've seen people spend
a long time (and harrass the wrong people) trying to trace the problem
with ns updates in the registry that turn out to be glue/authoritative
record inconsistencies.

The whole "glue" record concept is a necassary, but painful reality in
DNS (I'm not sure what how much BIND and other DNS software give
preference to records from an authoritative domain over a glue record in a
higher level domain, but the mixup between the two can make for some very
strange error).  I think there's a very valid argument for eliminating
replication nameserer IP information.  If it is clear that one NS record
per domain (at least a domain with nameservers in it) requires an IP
address then people can focus on keeping that one machine as a "glue"
between the TLD and the SLD and then use it to setup and maintain a large
number of secondary/tertiary NS's for the domain (which have no ip
addresses in the registry).  There is really only a need for one "glue"
between a TLD and an SLD (though a redundancy requirement, if dicated by
the registry, like NSI does, is not a bad idea).

Regards,
Sheer



Home | Date list | Subject list