Re: [RMONMIB] Overflow Counters (was comments on draft-ietf -rmonmib-dsmon-mib-02.txt)

[David Kaufman <dkaufman@desktalk.com>]

We also support both 64 bit counters and dual 32 bit counters.  With the 
spotty support of snmpv2c amongst both agent and management providers, it 
is problematic to rely exclusively on 64 bit counters in any environment 
particularly those where multiple management applications are being 
integrated.  Didn't we learn our lesson on this subject from the failed 
snmpv1.5 debate.
At 08:46 AM 10/2/00, lester@netscout.com wrote:


>Hi,
>
>On the surface it looks like a good idea to drop excess baggage by not
>having the overflow counters, but this could cause a source of confusion
>for the management stations.  In our applications we choose not to poll the
>64-bit counters, but rely on the two 32-bit counters (base + overflow).
>This way the application does not have to bother if the agent supports
>SNMPv2C or not.
>By dropping the 32-bit overflow counters, the management station will have
>to determine if the agent supports 64-bit counters or not and then poll
>only the OID's that the agent supports.  This can easily break the
>application when it encounters one agent implementation with 64-bit
>counters and one with  32-bit counters.
>
>In all I would say that  the MIBs should countinue to support the 32-bit
>overflow counter unless there is a huge advantage of not doing so.
>
>Lester D'Souza
>NetScout Systems, Inc.
>lester@netscout.com
>
>
>
>
>
>Andy Bierman <abierman@cisco.com>@ietf.org on 10/02/2000 11:07:00 AM
>
>Sent by:  rmonmib-admin@ietf.org
>
>
>To:   mibs@ops.ietf.org
>cc:   rmonmib@ietf.org
>Subject:  [RMONMIB] Overflow Counters (was comments on
>       draft-ietf-rmonmib-dsmon-mib-02.txt)
>
>
>[this was originally posted to the RMONMIB WG mailing list, but Bert asked
>  me to move the thread to this list. Sorry for the X-post, but many RMON WG
>  members are not on this list.]
>
>Hi,
>
>I would like to propose that it is time to stop duplicating Counter64
>objects in MIBs with a "Counter32/Overflow32 pair". It is time to
>require SNMPv2C support in applications to access 64-bit counters.
>
>I would like to remove all of the "overflow counters" from the DSMON MIB,
>(such as dsmonStatsInOvflPkts or dsmonStatsOutOvflOctets).  The 32-bit
>versions should remain, because DSMON could obviously be implemented on
>a device that doesn't need 64-bit counters.
>
>For devices that implement the Counter64 version of a counter object, the
>corresponding Counter32 version must also be present, and should equal the
>lower 32 bits of the 64-bit version at all times.
>
>Furthermore, it shall be the policy of the RMONMIB WG that no such
>"overflow counters" will be published in any future work.
>
>Any objections to changing the DSMON MIB and/or establishing this policy
>for the WG?
>
>Andy
>
>_______________________________________________
>RMONMIB mailing list
>RMONMIB@ietf.org
>http://www.ietf.org/mailman/listinfo/rmonmib
>
>
>
>
>_______________________________________________
>RMONMIB mailing list
>RMONMIB@ietf.org
>http://www.ietf.org/mailman/listinfo/rmonmib