On Wed, 21 Apr 2004, Jay W. Reffner wrote:

> I realize that the Squid SNMP agent is intended as an application SNMP
> agent and thus it doesn't report the MIB II tree. But by definition of
> the RFC's it should have, in my humble opinion, been implemented as an
> SNMP subagent (i.e. sagt), not a stand alone agent.

As I said in my prior response most master SNMP agents supports proxying
of SNMP requests to another SNMP agent. By using this you can easily make
the Squid SNMP agent a subagent of your master SNMP agent. This fully
satisfies the RFC and makes the full snmp MIB natural in the way it should
be. This is practically equivalent to using one of the more modern
subagent approaches such as AgentX or SMUX, just configured differently
and with the added benefit that the Squid agent can also be used
standalone with little or no effort.

In terms of resource usage there is no difference. A sub agent is
technically a master for it's own part of the tree. In fact adding
explicit subagent support requires more resources as the code for
registering with the master agent needs to be added to Squid, and in case
of AgentX a whole set of different SNMP packet formats need to be