Re: MIB module numbering for TBD MIB roots

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

On Sun, 20 Jan 2002, Bill Fenner wrote:
> On the versioning issue: I have a device that implements an old version
> of the SMON-MIB, but rooted under {rmon} as the RFC is.  Since objects
> had been rearranged between the draft that had been implemented and the
> RFC, I assumed that the bogus OIDs being returned were an implementation
> bug and so I spent quite a bit of time with technical support trying to
> figure out the issue.  Their response was that they had implemented
> a different version and sent it to me.

Yup, this is a problem, and it's the reason why stuff rooted under
mib-2 (at any level) should not appear in any WG document until it's
ready for publication.

> I guess I must be naieve to believe RFC 2578 when it says that
> MIBs being designed by working groups live under experimental,
> and then get moved to mgmt when they enter the standards track.

I vaguely recall this having been done in the past but it seems pretty
much to have been abandoned in favour of stuff like { mib-2 XXX }.
Frankly I don't see much benefit to the experimental tree, given that
the SMI imposes the same rules against OID reuse in all trees;  that
disallow rearrangements (as happened for the preliminary SMON version
you reported struggling with above) whenever a real OID appears in a
document (or at least in a document that receives wide circulation).

In my view, the { mib-2 XXX } is a good strategy to use in the early
stages, until the MIB structure stabilizes.  


Fri, 18 Jan 2002 10:15:49 -0800 Andy Bierman wrote:
> I think there is benefit in encouraging implementation before the
> very end of a standard MIB effort, especially since most
> MIBs stabilize fairly quickly, then bog down in cycles of reviews,
> clarifications, and delays.
>     
> WGs like RMONMIB that have their own branch under mib-2 have a clear
> advantage over WGs that have to wait until the end of the process
> to get a number.

There would be a real benefit in getting a number assigned
earlier in the process than is done now -- at the very least, it
should be possible to get a number as soon as something has been
approved for publication;  it should not be necessary to wait while
the document bubbles to the head of the RFC Editor's queue.  And if early
implementations (prior to approval for publication) are deemed desirable
or necessary, well, then that's what the experimental tree is for, and
it should be possible for a WG to get an node on the experimental tree
that can be delegated as it sees fit.

//cmh