Applicability Statement for DNS MIB Extensions
RFC 3197

Document Type RFC - Informational (November 2001; No errata)
Last updated 2013-03-02
Stream IETF
Formats plain text pdf html bibtex
Stream WG state (None)
Document shepherd No shepherd assigned
IESG IESG state RFC 3197 (Informational)
Consensus Boilerplate Unknown
Telechat date
Responsible AD (None)
Send notices to (None)
Network Working Group                                         R. Austein
Request for Comments: 3197                                 InterNetShare
Category: Informational                                    November 2001

             Applicability Statement for DNS MIB Extensions

Status of this Memo

   This memo provides information for the Internet community.  It does
   not specify an Internet standard of any kind.  Distribution of this
   memo is unlimited.

Copyright Notice

   Copyright (C) The Internet Society (2001).  All Rights Reserved.


   This document explains why, after more than six years as proposed
   standards, the DNS Server and Resolver MIB extensions were never
   deployed, and recommends retiring these MIB extensions by moving them
   to Historical status.

1. History

   The road to the DNS MIB extensions was paved with good intentions.

   In retrospect, it's obvious that the working group never had much
   agreement on what belonged in the MIB extensions, just that we should
   have some.  This happened during the height of the craze for MIB
   extensions in virtually every protocol that the IETF was working on
   at the time, so the question of why we were doing this in the first
   place never got a lot of scrutiny.  Very late in the development
   cycle we discovered that much of the support for writing the MIB
   extensions in the first place had come from people who wanted to use
   SNMP SET operations to update DNS zones on the fly.  Examination of
   the security model involved, however, led us to conclude that this
   was not a good way to do dynamic update and that a separate DNS
   Dynamic Update protocol would be necessary.

   The MIB extensions started out being fairly specific to one
   particular DNS implementation (BIND-4.8.3); as work progressed, the
   BIND-specific portions were rewritten to be as implementation-neutral
   as we knew how to make them, but somehow every revision of the MIB
   extensions managed to create new counters that just happened to
   closely match statistics kept by some version of BIND.  As a result,
   the MIB extensions ended up being much too big, which raised a number

Austein                      Informational                      [Page 1]
RFC 3197      Applicability Statement - DNS MIB Extensions November 2001

   of concerns with the network management directorate, but the WG
   resisted every attempt to remove any of these variables.  In the end,
   large portions of the MIB extensions were moved into optional groups
   in an attempt to get the required subset down to a manageable size.

   The DNS Server and Resolver MIB extensions were one of the first
   attempts to write MIB extensions for a protocol usually considered to
   be at the application layer.  Fairly early on it became clear that,
   while it was certainly possible to write MIB extensions for DNS, the
   SMI was not really designed with this sort of thing in mind.  A case
   in point was the attempt to provide direct indexing into the caches
   in the resolver MIB extensions: while arguably the only sane way to
   do this for a large cache, this required much more complex indexing
   clauses than is usual, and ended up running into known length limits
   for object identifiers in some SNMP implementations.

   Furthermore, the lack of either real proxy MIB support in SNMP
   managers or a standard subagent protocol meant that there was no
   reasonable way to implement the MIB extensions in the dominant
   implementation (BIND).  When the AgentX subagent protocol was
   developed a few years later, we initially hoped that this would
   finally clear the way for an implementation of the DNS MIB
   extensions, but by the time AgentX was a viable protocol it had
   become clear that nobody really wanted to implement these MIB

   Finally, the MIB extensions took much too long to produce.  In
   retrospect, this should have been a clear warning sign, particularly
   when the WG had clearly become so tired of the project that the
   authors found it impossible to elicit any comments whatsoever on the

2. Lessons

   Observations based on the preceding list of mistakes, for the benefit
   of anyone else who ever attempts to write DNS MIB extensions again:

   -  Define a clear set of goals before writing any MIB extensions.
      Know who the constituency is and make sure that what you write
      solves their problem.

   -  Keep the MIB extensions short, and don't add variables just
      because somebody in the WG thinks they'd be a cool thing to

   -  If some portion of the task seems to be very hard to do within the
      SMI, that's a strong hint that SNMP is not the right tool for
      whatever it is that you're trying to do.
Show full document text