Re: ifNumber and its ilk considered harmful

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

On Thu, 6 Feb 2003, Mike MacFaden wrote:
> 
> ----- Forwarded message from Margaret Wasserman <mrw@windriver.com> -----
> 
> >From: Margaret Wasserman <mrw@windriver.com>
> >Subject: Re: Please add back FIB counter to 2096 update
> >
> >Hi Mike,
> >
> >I hear conflicting things from various MIB experts about
> >whether or not the "number" variables are useful.
> >
> >They don't represent the number of objects within the
> >view, and they also are hard (maybe impossible?) to
> >implement on distributed systems such as AgentX.
> >
> >So, I thought there were sufficiently out-of-favor that
> >we shouldn't continue with them...
> >
> >Apparently you disagree, though?  Is this something that
> >the MIB experts could work out between themselves and
> >reach a consistent opinion regarding?
> >
> ----- End forwarded message -----
> 
> Comments?
> 
> Have any recent reviews disapproved of using scalars like ifNumber?

Not from me, certainly.  Indeed, I agree that the FIB counter
inetCidrROuteNumber should be re-instated into 2096-update,
and said so in my recent comments on the ipv6 list.  The
inetCidrRouteTable can be HUGE (100Ks of entries), and it's
obviously not efficient (and probably not even feasible) for
a management station to count the number of entries by
retrieving the whole table.

I have long understood that the AgentX folks dislike summary objects
objects such as ifNumber that count the number of rows in a table
where different rows may be in different subagents.  But
implementation difficulty for agents is only part of the story. The
other part is that these summary counters (reportedly) can be very
useful to management applications.  There does not seem to be a
general, one-size-fits-all "yes summary counters are good" or
"no summary counters are bad" type of guideline, but only difficult
engineering tradeoffs.

In MIB reviews I frequently encounter technologies with which I am
not familiar.  In such cases I almost always defer to the judgement
of the folks that know the technology when it comes to engineering
tradeoffs.  What I would be likely to do if I saw a summary counter
in a MIB module I was reviewing would be to alert the authors about
the tradeoffs involved, and then let the decision be up to them.

//cmh