Re: Missing normative references to modules containing importeditems (was: FW: Malloc MIB)

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

On Wed, 29 Jan 2003, Wijnen, Bert (Bert) wrote:

> Mmmm... I wonder why this happens more than it should.
> 
> - I guess one reason is that we've not been checking that
>   too rigurously

I've certainly been guilty of that;  as Bert is well aware, I
muffed it on two recent documents (OPT-IF-MIB and MALLOC-MIB).

The main problem for me (at least in the case of the MALLOC-MIB) was
that this item was not on my MIB review checklist when I did the
original review;  in fact it was completely off my radar until
someone brought it up on this list.  It's now in two places in the
MIB review guidelines document.  One is in Section 3.5, "References
Sections" --

   The standard MIB boilerplate available at
   http://www.ops.ietf.org/mib-boilerplate.html includes lists of
   normative and informative references that MUST appear in all
   specifications that contain MIB modules.  If items from other MIB
   modules appear in an IMPORTS statement in the Definitions section,
   then the specifications containing those other MIB modules MUST be
   included in the list of normative references.

and the other is in Appendix A (checklist):

   5.) References -- verify [ ... ] that all MIB modules containing
   imported items are cited as normative references [ ... ]

The problem was compounded because of the reorganization of references
that took place after the publication of the new SNMPv3 RFCs:  in the
OPT-IF-MIB RFC 2571 was removed from the normative list and was not
automatically replaced by RFC 3411 because the boilerplate no longer
requires it.  Unfortunately I forgot to check the IMPORTS list because
I was focussed purely on the changes, and I did not catch the omission
until a couple of weeks after the new draft was submitted :-(

> - Another is that it is not always clear since the 
>   reference does [not] exist.
>   Several years back I saw things like:
> 
>      IMPORT
>      SnmpAdminString 
>      FROM   SNMp-FRAMEWORK-MIB  -- RFC xxxx [RFCxxxx]
> 
>   But some one objected to that... and since then I do not
>   see it much (if at all). If we started to recommend that
>   approach, then the normative references would more easily
>   show up.

Speaking for myself, I don't think that's an absolute necessary
now that the check is in the references (and of course having
muffed it twice I'm not likely to forget again -- or at least so
I hope :)

However, IF the consensus of this list is that we should put
RFC numbers in ASN.1 comments above, then I'll be happy to add
that guideline to the section that deals with the IMPORTS
statement.  There are some cases were a reference appears ONLY
because of an imported item -- e.g., in the OPT-IF-MIB the only
need for RFC 3411 was for the import of SnmpAdminString -- and
it could be useful to ensure that each reference is tied to at
least one citation.  On the other hand, it's probably deadweight
in cases  where other citations exist (as will surely be the case
for SNMPv2-SMI, SNMPv2-TC, and SNMPv2-CONF).

Mike