RE: Updates to rfc2011, rfc2096 others

["C. M. Heard" <heard@pobox.com>]

On Thu, 29 Jan 2004, Wijnen, Bert (Bert) wrote:
> When this work first started, I had not expected that all
> existing objects in 2096 (for example) would be deprecated and
> that we would end up will only new current objects.

Well, exactly the same thing happened in the transition from RFC
1354 to RFC 2096, except that in that case all the old objects were
obsoleted rather than deprecated.  And there really wasn't any other
possible outcome for the update of RFC 2096 if the following advice
from RFC 3291 was to be followed:

  MIB modules of the first type SHOULD provide object definitions
  (e.g., tables) that work with all versions of IP.  In particular,
  when revising a MIB module which contains IPv4 specific tables,
  it is suggested to define new tables using the textual conventions
  defined in this memo which support all versions of IP.  The status
  of the new tables SHOULD be "current" while the status of the old
  IP version specific tables SHOULD be changed to "deprecated".  The
  other approach of having multiple similar tables for different IP
  versions is strongly discouraged.

Since the only things in RFC 2096 (apart from the obsolete objects
that were originally in RFC 1354) were one IPv4-specific table and
one scalar to count the number of valid entries in that table, the
above advice pretty much dictates that at the end of the day the
only current objects would be new ones.

Is this a perhaps case where some apparently reasonable SHOULDs in
RFC 3291 had unintended (or at least unanticipated) consequences?

//cmh