applying the model

[Jeff Yarnell <[e-mail]>]

After posting some questions to the list a few weeks ago, I 
have been trying to apply the opstat model to standard MIBs
other than MIB-II. The responses to my questions confirmed
that the model is biased towards polling routers for MIB-II
statistics.

Please don't misunderstand my comments. I think the work that
has been so far is great.  I'm just trying to apply it and,
perhaps, stretch it a little.  I'd like to see the opstat
work continue and become more robust.

I have been considering how the model and specifically the
storage format may be applied to the RMON MIB. I have found
some difficulties and am interested in other's opinions as how
the problems may be solved.

Problem:
The opstat model's means of identifying variables polled
is to put them in the tag-table.  According to RFC1404, "Variable names 
are the fully qualified Internet MIB names."  The examples given show 
things like, "ifInOctets."  This works well when the variable being 
polled is either a scalar, or has a one-to-one correspondence with the 
number of interfaces on the polled device (e.g. "ifInOctets"). However, 
this doesn't work for some tables. Consider the RMON host table. The 
table has entries for each physical address seen by the probe and is 
indexed by the physical address. I don't see how the current storage 
format could accommodate polling objects in such a table. If the
device section's tag table includes "hostInPkts" which host is it?
How does the server indicate that it's polling hostInPkts for all
hosts in the table?

It seems to me that either the device section must be modified such
that variable-names include the instance portion of the variable polled
or the data section must have instance identification associated
with each data-value

Interested in your thoughts,
Jeff Yarnell
Kaspia Systems, Inc.
503-620-5571