Re: FW: draft-ietf-rap-feedback-fr-pib-05.txt submission

[Juergen Schoenwaelder <schoenw@ibr.cs.tu-bs.de>]

>>>>> Bokaemper, Martin writes:

Martin> Unchanged: - Usage32/64 - see todays discussion on the list

                If there is the possibility that the maximum Usage32
                value of 2^32-1 is exceeded during the lifetime
                of the Usage32 object, the larger Usage64 type
                should be used.

                The lifetime of the Usage64 object should be defined
                in a way that ensures the maximum Usage64 value of
                2^64-1 is never exceeded.

I do not really know how a PIB author should make such a decision.
There are two factors here - the lifetime you assume and the frequency
of the events you are counting. In the MIB space we have learned that
technology changes in just a few years can totally change how long it
takes to fill up 32 bit counters.

Now assume that I have implemented a Usage32 object and the event
frequency changes such that I reach the upper limit - what is going to
happen? Is your text saying that in case I reach the end of my number
space, I have to end the lifetime of the object?

Or is the text trying to tell me that PIB authors SHOULD choose the
right UsageXX type (which in most cases will just be Usage64) which
avoids rollovers but the type itself does rollover just like a normal
SMIv2 CounterXX type?

Perhaps just some rewording would help to distinguish between (a) how
the type behaves and (b) how the type should be used by PIB authors.

/js

-- 
Juergen Schoenwaelder    <http://www.informatik.uni-osnabrueck.de/schoenw/>