Re: snmpEngineBoots

Hi -

> From: "Ann Lo" <[email protected]>
> To: "C. M. Heard" <[email protected]>
> Cc: <[email protected]>
> Sent: Monday, May 05, 2003 1:24 PM
> Subject: Re: snmpEngineBoots
...
> Thanks for your information. I guess that this definition has originated
> from the Unix system where the SNMP protocol can be operational when the
> SNMP agent is down?

No.

> In many embedded systems, snmpEngineBoots would then start from 1 as these
> systems would not respond to any SNMP requests until the SNMP agent is up
> and running.

No.  The snmpEngineBoots variable needs to persist, regardless of the
operating system.  Whenever an SNMP engine starts up, the value needs to be
incremented.  Consequently, if you ever want to see it on the wire with a
value of 1, theinitial value (prior to the first time the SNMP engine is ever
started) must be zero.  (This is just what Mike explained.)

Randy

> Regards,
> Ann
>
>
> "C. M. Heard" wrote:
>
> > On Fri, 2 May 2003, Ann Lo wrote:
> > > Should snmpEngineBoots be initialized to 0 or 1? rfc3414 (USM)
> > > says 0.
> >
> > That's the correct initial value.
> >
> > > rfc3411 says 1.
> >
> > More precisely, it says that's the minimum value that can be
> > returned in respose to a Get-Request-PDU.
> >
> > > To be really consistent with both rfc's, the implemention could be as
> > > follows:
> > >
> > >    * Set snmpEngineBoots to 0 when the SNMP agent is first installed.
> > >    * Do not provide a MIB instance to snmpEngineBoots when it is 0.
> >
> > Note that the engine _cannot_ return any value until after it is
> > booted at least one, so the second step is automatic.
> >
> > I seem to recall that this discussion has taken place on this list
> > before.
> >
> > //cmh
>