Re: What's needed

[Bob Friesenhahn <[e-mail]>]

According to Karl Auerbach, Empirical Tools and Technologies, 408/427-5280:
> 
>  > The way MIBs are documented now, its very hard for an NMS vendor to
>  > do useful things with the data.
> 
> I certainly agree with you that much better MIB definitions and
> formalities would be a great boon.  Not only do we need better
> definitions, but also inter-variable relationship assertions, action
> parameter/result-value definitions, and perhaps even pathology
> recognition (or at least pathology suggesting) expressions.

Yes, yes!  I would like to see a solution to the classical
bomb/bomb-bay door problem imbedded somehow in the MIB for both
agents and NMSs to use.  If this support was there, then much of
the development load could be removed from agent and NMS developers.

> If one *really* wants to build a useful NMS, one need not simply
> bemoan the lack of machine readible, formalized definitions embedded
> in MIBs.
> 
> One could just pick up the phone or hit the e-mail keyboard and ask the
> developer.

This presumes that the developer really knows what normal values
are for their equipment.  Typically equipment is stressed at maximum
load and then it is taken to some customer site for real testing.
Normal values are likely to vary from network to network.  The NMS
really has to be able to adapt and tailor itself to what the users
consider to be "normal" at a particular site.

Bob
-- 
Bob Friesenhahn, [phone]
Email:		[e-mail]
US mail:	4333 Harvest Hill, Carrollton, TX 75010