[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: BCP and Notifications (was: snmpconf FW: MIBs for events andnotification logging)


It looks like you are missing the point, and it certainly doesn't
help your argument to use terms like "red herring" and "angles
on heads of pins". Now you change from "reliability" to
"verifiablity". If you want to say something like
"this approach will work in these situations ..., and not
work in these situations ..., and the first situations
occur x% of the time, and the second situations occur 100-x%
of the time with a variation of %y, or due to z. And the
result of the failure of the mechanism to support the situation
is q." Then great, I'll defend your position (if you show
me how you got to the numbers). However, to make the broad
statements that I know to not be accurate results in me
sounding an alarm. From my reading and personal experience,
when major disasters occur, it is generally not when there
is a single failure, but when one or more failures result
in others that cassade and cause "meltdown".

On Fri, 2 Nov 2001, Jon Saperia wrote:
> The important point here is not reliability, by verifiability.  Even
> when we have logs on the managed device, that does not guarantee that
> the message will get through. Networks go down sometimes for a long
> periods of time. The idea is: did the message get through in some finite
> period of time, and if not what is done? Debating 'reliability' is a red
> herring. In the end, with reasonable retry implementations, the message
> gets through enough of the time to be helpful when using
> INFORMS. Discussion beyond that falls into the realm of how many angels
> can dance on the head of a pin.

/david t. perkins