Re: Questions on HDLC MIB

[Mark Therieau <[e-mail]>]

	>> I also agree that the "nearest analog" solution is a good
	>> compromise.  But, in a case where no approximation can be made,
	>> returning zero for an error counter is misleading.

	What is your preferred solution?

	Fred

In my case I will undertake the "nearest analog" solution.

My preferred solution is the one that allows the majority of HDLC
implementations to support this MIB.  But, the MIB must be effective in
instrumenting HDLC.  These two goals may be conflicting.

Since there is no existing standard on HDLC management objects, current
HDLC implementations almost necessarily support a varied range of
instrumentation capabilities.  The set of supported objects for a given 
implementation may be deficient in certain areas.

The real question is: what subset of the objects that are necessary and 
sufficient to instrument HDLC?

In the absence of an existing set of standard objects to instrument HDLC,
I yield to the better minds (and greater experience) in this group the job
of defining the core set of objects.

I yield the point that if an object is considered to be a part of the core
set, the inability of some implementations to support the object doesn't
mean that the object no longer belongs in the core set.

I would also argue that if an implementation can only partially support an
object, (by restricted range of values or "nearest analog" approximation) that
implementation is still conformant.

markt


----- End Included Message -----