Re: Object Identifier refinements.

["Randy Presuhn" <randy_presuhn@mindspring.com>]

Hi -

> From: "Michael Kirkham" <mikek@muonics.com>
> To: <sming@ops.ietf.org>
> Sent: Saturday, January 11, 2003 8:38 PM
> Subject: Object Identifier refinements.
...
> I would like to suggest that object identifiers should be able to be
> refined in SMIng in order to specify a limit to the number of
> subidentifiers that can be present when an object identifier type object
> is used as an index.  Otherwise, it is difficult for a validation tool to
> determine the author's intentions -- ie., whether they expect the limit to
> be the maximum number that will fit given the object's depth in OID space
> and the other indices, or whether the author is forgetting to take the
> 128-subidentifier limit into account and is defining something illegal.
> Allowing them to have refinements would allow a validation tool to a
> potential conflict if an OID index has no refinement or know definitively
> if there is a conflict if there is a refinement.
...

I think the 128 subidentifier limit was a mistake, and that adding size
constraints to OBJECT IDENTIFIER would only compound it.
The definer of an object cannot know how many levels of indirection might
be involved in references to that object.  All the stuff that might be referenced
by VACM is an example of this, as is snmpNotifyFilterSubtree in RFC 3413.

Randy