Dave, David,
Thanks a lot for your detailed responses.
Cheers,
Dipesh
"David T. Perkins" <[EMAIL PROTECTED]> wrote:
HI,
A "better solution" is "better MIB design". And of course
some of us wished that the error code "tooComplex" was
present.
But seriously, a MIB designer can make a huge difference.
Of course, if you are implementing a MIB that was designed
by someone that had no idea of the cost to code the design,
you are sort of caugth.
Also, sometimes it is more appropriate to use another
protocol instead of SNMP.
On Tue, 20 Sep 2005, Dave Shield wrote:
> On Mon, 2005-09-19 at 11:01 -0700, Dipesh Gorashia wrote:
>
> > Question:
> > While proccessing varbinds "simultaneousluy", the
> > above consistency check
> >
> > --> should be made for each variable in varbind only
> > against existing configuration on the system?
> >
> > --> should the varia bles in varbinds be validated for
> > consistency also against each other?
>
> I think this validation *MUST* take into account the
> other varbinds in the assignment. There could easily
> be situations where both the initial and final
> configurations are consistent - but any "partial
> update" isn't.
>
> For example, consider a SNMP-controlled (3-ball)
> juggling robot, that can only hold one ball in each
> hand, operating in a very confined space (so that only
> one ball can be in the air at any one time!).
>
> That could be monitored and updated using SNMP commands
> such as:
>
> $ snmpwalk juggler ballLocation
> ballLocation.1 = leftHand
> ballLocation.2 = rightHand
> ballLocation.3 = inTheAir
>
> $ snmpset juggler ballAction.1 = throwFromLeft
> ballAction.2 = passRightToLeft
> ballAction.3 = catchInRight
>
> $ snmp walk juggler ballLocation
> ballLocation.1 = inTheAir
> ballLocation.2 = leftHand
> ballLocation.3 = rightHand
>
>
> That's a perfectly reasonable (if somewhat contrived!)
> example - although the robot might have problems in
> stopping cleanly :-).
> But if you checked the consistency of any of the three
> updates individually against the previous configuration,
> the SET request would be rejected as inconsistent.
>
> It's only when all three updates are taken together,
> that the assignment becomes acceptable.
>
>
> The most common technique is probably to apply *all*
> the assignments to some internal representation of the
> system, then check the resulting configuration for
> consistency. If this check fails, then restore the
> internal representation to its previous state.
> If this check succeeds, then apply these changes to
> the ex ternal system (if relevant).
>
> This is not always possible, of course - but that's
> a useful basic working principle.
>
> Dave
>
Regards,
/david t. perkins
I thought I wanted a career; turns out I just wanted a salary.
Yahoo! for Good
Click here to donate to the Hurricane Katrina relief effort.
