Discontinuity timer for TCP-MIB counters

Hi Folks,

This point came out of the IESG review of the TCP-MIB, and possibly
requires wider discussion. Hence this mail to all the aliases.

The current update to TCP-MIB (draft-ietf-ipv6-rfc2012-update-05.txt)
does not define any special discontinuity objects, for a case
when the counters in the mib experience a discontinuity. There is
an implicit assumption that this would be represented by sysUpTime
(SNMPv2-MIB).

The IESG review suggested that we provide a discontinuity timer
for the counters in the mib - could be via sysUpTime or a new
discontinuity object. But we state it either way. There are issues
to going either way:

1. sysUpTime: If any Counter, that uses sysUpTime as the
    discontinuity indicator, experiences a discontinuity, then
    it must reset sysUpTime (basically means a warm (re-)start).
    Therefore, all *other* counters that also use sysUpTime are
    going to look and be treated as if they also had a discontinuity.
    This may have performance implications, in terms of network
    management stations generating additional queries for these
    (unaffected) counters as well.

2. Add a new object for indicate discontinuity: This would
    necessitate adding a new (perhaps) unnecessary object, if
    sysUpTime can serve the purpose. In addition, systems would
    have an additional overhead of maintaining this information.

Hence the question to folks here is:

1. Are there any known systems that can experience a discontinuity
    in the TCP counters without affecting the network management
    portion of the system. For e.g. are there systems that can
    restart the TCP portion of the system without restarting the
    network management portion of the system (or the whole system)?

2. Does anyone see any other case in which we might need to/want to
    maintain a discontinuity timer in the TCP-MIB?

My assumption is that there might not be any systems that need an
explicit discontinuity timer (but I could be wrong). Hence, would
appreciate responses.

Thanks,
Rajiv.

--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------