Re: ifNumber and its ilk considered harmful

[Mike MacFaden <mrm@riverstonenet.com>]

On Fri, Feb 07, 2003 at 01:52:02PM -0800, Andy Bierman wrote:
>>An alternative I proposed to Margaret was to follow the model
>>used in rfc2021 for HUGE tables: namely using two counters
>>to track inserts and deletes. 
>
>This doesn't make any sense.  If an implementation utilizing
>subagents can support two counters (inserts and deletes) then
>it can easily support a single gauge like ifNumber.  It just
>implements ifNumber internally as two counters and returns
>inserts - deletes when asked for ifNumber.

Margaret offered two arguments against using table counters.
The second involved the use of views such that the number
might not reflect the number of rows visible. That is
why I suggested an alternate approach.

I've never had to use agent-x but
I am beginning to think that any *summary* statistic 
must be hard for an agent-x implementation.

It is a pity the revised standard does not
provide any of the useful counters found in many
different vendor implementations of the cli command: 
'ip show route summary' 

Mike